The Imperial CollegeLecturesinPetroleumEngineeringVolume 3

The Imperial College Lectures in
PETROLEUM ENGINEERING
Topics in Reservoir Management
Volume
Q0082_9781786342843_TP.indd 1 11/8/17 11:33 AM

b2530 International Strategic Relations and China’s National Security: World at the Crossroads
This page intentionally left blank
b2530_FM.indd 6 01-Sep-16 11:03:06 AM

The Imperial College Lectures in
PETROLEUM ENGINEERING
Topics in Reservoir Management
Volume
Deryck Bond
Kuwait Oil Company, Kuwait
Samuel Krevor
Imperial College London, UK
Ann Muggeridge
David Waldren
Petroleum Consulting and Training (PCT) Ltd, UK
Robert Zimmerman
World Scientific
NEW JERSEY • LONDON • SINGAPORE • BEIJING • SHANGHAI • HONG KONG • TAIPEI • CHENNAI • TOKYO
Q0082_9781786342843_TP.indd 2 11/8/17 11:33 AM

Published by
World Scientific Publishing Europe Ltd.
57 Shelton Street, Covent Garden, London WC2H 9HE
Head office: 5 Toh Tuck Link, Singapore 596224
USA office: 27 Warren Street, Suite 401-402, Hackensack, NJ 07601
Library of Congress Cataloging-in-Publication Data

Names: Muggeridge, Ann.
Title: Topics in reservoir management / by Ann Muggeridge (Imperial College London, UK),
Sam Krevor (Imperial College London, UK), Robert Zimmerman
(Imperial College London, UK), Deryck Bond (Kuwait Oil Company, Kuwait),
David Waldren (Petroleum Consulting and Training (PCT) Ltd, UK).
Description: [Hackensack] New Jersey : World Scientific, 2017. | Series: The Imperial College
lectures in petroleum engineering ; volume 3 | Based on lectures that have been given in the
world-renowned Imperial College Masters Course in Petroleum Engineering.
Identifiers: LCCN 2016055436 | ISBN 9781786342843 (hc : alk. paper)
Subjects: LCSH: Oil reservoir engineering. | Oil fields--Production methods. |
Petroleum industry and trade--Management.
Classification: LCC TN870.57 .M84 2017 | DDC 622/.3382--dc23
LC record available at https://lccn.loc.gov/2016055436
British Library Cataloguing-in-Publication Data

A catalogue record for this book is available from the British Library.
Copyright © 2018 by World Scientific Publishing Europe Ltd.

All rights reserved. This book, or parts thereof, may not be reproduced in any form or by any means,
electronic or mechanical, including photocopying, recording or any information storage and retrieval
system now known or to be invented, without written permission from the Publisher.
For photocopying of material in this volume, please pay a copying fee through the Copyright Clearance
Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to photocopy
is not required from the publisher.
Desk Editors: Dipasri Sardar/Mary Simpson/Shi Ying Koe/Jennifer Brough
Typeset by Stallion Press

Email: enquiries@stallionpress.com
Printed in Singapore
Dipa - The Imperial College Lectures - V3.indd 1 26-07-17 3:51:51 PM

August 11, 2017 10:20 Topics in Reservoir Management - 9in x 6in b2764-fm page v
Preface
This book is the third volume of a set of lecture notes based on the
Master of Science course in Petroleum Engineering that is taught
within the Department of Earth Science and Engineering at Imperial
College London. The Petroleum Engineering MSc is a one-year course
that comprises three components: (a) a set of lectures on the different
topics that constitute the field of petroleum engineering, along with
associated homework assignments and examinations; (b) a group field
project in which the class is broken up into groups of about six
students, who then use data from an actual reservoir to develop the
field from the initial appraisal based on seismic and geological data,
all the way through to eventual abandonment; and (c) a 14-week
individual project, in which each student investigates a specific
problem and writes a small “thesis” in the format of an SPE paper.
The Petroleum Engineering MSc course has been taught at
Imperial College since 1976, and has trained over a thousand
petroleum engineers. The course is essentially a “conversion course”
that aims to take students who have an undergraduate degree in
some area of engineering or physical science, but not necessarily
any specific experience in petroleum engineering, and train them
to the point at which they can enter the oil and gas industry as
petroleum engineers. Although the incoming cohort has included
students with undergraduate degrees in fields as varied as physics,
mathematics, geology, and electrical engineering, the “typical” stu-
dent on the course has an undergraduate degree in chemical or
v
August 11, 2017 10:20 Topics in Reservoir Management - 9in x 6in b2764-fm page vi
vi Topics in Reservoir Management
mechanical engineering, and little if any prior exposure to petroleum

engineering. Although some students do enter the course having
had some experience in the oil industry, the course is intended to
be self-contained, and prior knowledge of petroleum engineering or
geology is not a prerequisite for any of the lecture modules.
The complete set of lecture notes will eventually consist of about
a half-dozen different volumes, covering topics such as petroleum
geology, fluid flow in porous media, well test analysis, reservoir
engineering, and core analysis. The present volume contains chapters
on four topics that fall under the rubric of “reservoir management”:
rock properties, enhanced oil recovery, reservoir simulation, and
history matching. Each chapter has been written by a lecturer who
is either an academic based at Imperial College, or a practitioner
working in the oil industry, and who has taught these and other
modules within the Imperial College MSc course for many years.
Although the volumes are lecture notes, and consequently do not aim
to achieve encyclopaedic coverage, or to contain extensive reference
lists, taken as a whole they contain the basic knowledge needed to
prepare students to work as reservoir engineers in the oil and gas
industry.
Robert W. Zimmerman
Imperial College London
July 2017
August 23, 2017 14:34 Topics in Reservoir Management - 9in x 6in b2764-fm page vii
About the Authors
Deryck Bond holds BSc and PhD degrees in Physics from Imperial
College London. He has worked in research, reservoir engineering and
petroleum engineering functions for a number of major and mid-sized
oil companies and as an independent consultant. He has contributed
to the Petroleum Engineering MSc program at Imperial College. He
is currently a Senior Consultant working on the Greater Burhan field
for Kuwait Oil Company.
Samuel C. Krevor is a Senior Lecturer in the Department of

Earth Science and Engineering at Imperial College London. His
research group investigates the physics and chemistry of processes
of multiphase flow in engineered and natural geological systems. He
received his BSc, MSc, and PhD in Environmental Engineering from
Columbia University.
Ann H. Muggeridge holds the Chair in Reservoir Physics and EOR

in the Department of Earth Sciences and Engineering at Imperial
College London. Her research interests are focussed on the numerical
modelling and upscaling of reservoir flows with a particular emphasis
on enhanced oil recovery processes (including miscible gas injection,
low salinity waterflooding, polymer flooding, steam injection and
Vapour Extraction (VAPEX)) and the impact of geological hetero-
geneity on those processes. She is also interested in the mixing of
reservoir fluids after reservoir filling but before production as an
vii
August 11, 2017 10:20 Topics in Reservoir Management - 9in x 6in b2764-fm page viii
viii Topics in Reservoir Management
indicator of reservoir compartmentalization. She has a BSc (Hons) in

Physics from Imperial College and a DPhil in Atmospheric Physics
from the University of Oxford. She worked for 5 years at BP and
4 years at SSI (UK) Ltd as a reservoir engineering and reservoir
simulation specialist before joining Imperial College. She is a member
of the EAGE and the SPE, chairing the organizing committee for the
EAGE IOR Symposium 2017 and 2015 and sitting on the organizing
committee for the SPE Reservoir Simulation Symposium 2017. She is
also an InterPore Council member and sits on the Scientific Advisory
Committee for the National IOR Centre of Norway.
David Waldren is director of Petroleum Consulting and Training

Limited and a Visiting Professor to the Centre for Petroleum Studies,
Department of Earth Science and Engineering, Imperial College
London. He has been a consulting reservoir engineer for over 25 years,
previously having been with BNOC, Intercomp, and IPEC. He is
recognised worldwide for his reservoir management and simulation
advice. He has a PhD in particle physics from Liverpool University.
Robert W. Zimmerman obtained a BS and MS in mechanical

engineering from Columbia University, and a PhD in rock mechanics
from the University of California at Berkeley. He has been a lecturer
at UC Berkeley, a staff scientist at the Lawrence Berkeley National
Laboratory, and Head of the Division of Engineering Geology and
Geophysics at the Royal Institute of Technology (KTH) in Stock-
holm. He is the Editor-in-Chief of the International Journal of Rock
Mechanics and Mining Sciences, and serves on the Editorial Boards
of Transport in Porous Media and the International Journal of Engi-
neering Science. He is the author of the monograph Compressibility
of Sandstones (Elsevier, 1991), and co-author, with JC Jaeger and
NGW Cook, of Fundamentals of Rock Mechanics (4th edn., Wiley-
Blackwell, 2007). He is currently Professor of Rock Mechanics at
Imperial College, where he conducts research on rock mechanics and
fractured rock hydrology, with applications to petroleum engineering,
underground mining, carbon sequestration, and radioactive waste
disposal.
August 11, 2017 10:20 Topics in Reservoir Management - 9in x 6in b2764-fm page ix
Contents
Preface v
About the Authors vii
Chapter 1. Introduction to Rock Properties 1

Robert W. Zimmerman
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Porosity and Saturation . . . . . . . . . . . . . . . . . . . . 2
1.2.1 Definition of Porosity . . . . . . . . . . . . . . . . . 2
1.2.2 Heterogeneity and “Representative Elementary
Volume” . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2.3 Saturation . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3 Permeability and Darcy’s Law . . . . . . . . . . . . . . . . 7
1.3.1 Darcy’s Law . . . . . . . . . . . . . . . . . . . . . . . 7
1.3.2 Units of Permeability . . . . . . . . . . . . . . . . . 11
1.3.3 Relationship between Permeability and
Pore Size . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.3.4 Permeability of Layered Rocks . . . . . . . . . . . 14
1.3.5 Permeability Heterogeneity . . . . . . . . . . . . . 17
1.4 Surface Tension, Wettability and Capillarity . . . . . . . 18
1.4.1 Surface Tension . . . . . . . . . . . . . . . . . . . . . 18
1.4.2 Capillary Pressure . . . . . . . . . . . . . . . . . . . 21
1.4.3 Contact Angles . . . . . . . . . . . . . . . . . . . . . 22
ix
August 11, 2017 10:20 Topics in Reservoir Management - 9in x 6in b2764-fm page x
x Topics in Reservoir Management
1.4.4 Capillary Rise . . . . . . . . . . . . . . . . . . . . . . 25

1.4.5 Oil–Water Transition Zone . . . . . . . . . . . . . 27
1.4.6 Leverett J-Function . . . . . . . . . . . . . . . . . . 30
1.5 Two-Phase Flow and Relative Permeability . . . . . . . 32
1.5.1 Concept of Relative Permeability . . . . . . . . . 32
1.5.2 Irreducible Saturations . . . . . . . . . . . . . . . . 34
1.6 Electrical Resistivity . . . . . . . . . . . . . . . . . . . . . . 35
1.7 Fluid and Pore Compressibility . . . . . . . . . . . . . . . 41
1.7.1 Fluid Compressibility . . . . . . . . . . . . . . . . . 41
1.7.2 Pore Compressibility . . . . . . . . . . . . . . . . . 42
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
Questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
Chapter 2. Introduction to Enhanced Recovery

Processes for Conventional Oil
Production 47
Samuel C. Krevor and Ann H. Muggeridge
2.1 Introduction: Definition, Techniques and the Global

Role of EOR . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.1.1 The Aims of this Module . . . . . . . . . . . . . . 47
2.1.2 Definitions and Techniques . . . . . . . . . . . . . 48
2.1.3 The Role of EOR in Current and Future Global
Oil Production . . . . . . . . . . . . . . . . . . . . . 50
2.2 Enhancing the Recovery Factor . . . . . . . . . . . . . . . 53
2.2.1 The Recovery Factor . . . . . . . . . . . . . . . . . 53
2.2.2 Limits on Microscopic Displacement
Efficiency . . . . . . . . . . . . . . . . . . . . . . . . . 55
2.2.3 Limits on Macroscopic Displacement
Efficiency . . . . . . . . . . . . . . . . . . . . . . . . . 59
2.3 Gas Injection . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
2.3.1 Phase Equilibrium for Gas Drives . . . . . . . . . 68
2.3.2 Transport in Two-Phase Multi-Component
Systems . . . . . . . . . . . . . . . . . . . . . . . . . . 79
August 11, 2017 10:20 Topics in Reservoir Management - 9in x 6in b2764-fm page xi
Contents xi
2.3.2.1Review of the Buckley–Leverett

solution . . . . . . . . . . . . . . . . . . . . 80
2.3.2.2 Two-component, two-phase
displacement . . . . . . . . . . . . . . . . 82
2.3.3 Fractional Flow Theory and Water Alternating
Gas Injection . . . . . . . . . . . . . . . . . . . . . . 85
2.3.4 Three-Phase Flow . . . . . . . . . . . . . . . . . . . 90
2.3.4.1 Fractional flow theory applied to
three-phase flow . . . . . . . . . . . . . . 90
2.3.4.2 Models for three-phase relative
permeability . . . . . . . . . . . . . . . . . 91
2.4 Chemical EOR: Polymers and Low Salinity
Flooding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
2.4.1 Polymer Flooding . . . . . . . . . . . . . . . . . . . 93
2.4.2 Low Salinity Water Flooding . . . . . . . . . . . . 97
2.5 Practical Considerations . . . . . . . . . . . . . . . . . . . . 100
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
Chapter 3. Numerical Simulation 109

Dave Waldren
3.1 Reservoir Models . . . . . . . . . . . . . . . . . . . . . . . . . 109

3.1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . 109
3.1.2 Physical Models . . . . . . . . . . . . . . . . . . . . 110
3.1.2.1 Analogue models . . . . . . . . . . . . . . 110
3.1.2.2 Comprehensive models . . . . . . . . . . 111
3.1.2.3 Elemental models . . . . . . . . . . . . . 111
3.1.3 Mathematical Models . . . . . . . . . . . . . . . . . 112
3.1.4 Numerical Models . . . . . . . . . . . . . . . . . . . 112
3.1.5 Models as Comparative Tools . . . . . . . . . . . 113
3.2 Equations and Terminology . . . . . . . . . . . . . . . . . . 114
3.2.1 Mass Conservation . . . . . . . . . . . . . . . . . . . 114
3.2.2 Darcy’s Law . . . . . . . . . . . . . . . . . . . . . . . 115
3.2.3 Diffusivity Equation . . . . . . . . . . . . . . . . . . 117
3.2.4 Finite Difference . . . . . . . . . . . . . . . . . . . . 119
August 11, 2017 10:58 Topics in Reservoir Management - 9in x 6in b2764-fm page xii
xii Topics in Reservoir Management
3.2.5 Implicit and Explicit Formulation . . . . . . . . . 121

3.2.6 Dispersion and Weighting . . . . . . . . . . . . . . 125
3.2.7 Nonlinearity and Outer Iterations . . . . . . . . . 126
3.2.8 Linear Solvers . . . . . . . . . . . . . . . . . . . . . . 127
3.3 Buckley–Leverett Displacement . . . . . . . . . . . . . . . 131
3.4 Reservoir Models . . . . . . . . . . . . . . . . . . . . . . . . . 134
3.4.1 Model Components . . . . . . . . . . . . . . . . . . 134
3.4.1.1 Reservoir description . . . . . . . . . . . 134
3.4.1.2 Initialisation . . . . . . . . . . . . . . . . 134
3.4.1.3 Model control . . . . . . . . . . . . . . . . 134
3.4.1.4 Production data . . . . . . . . . . . . . . 135
3.4.1.5 Output . . . . . . . . . . . . . . . . . . . . 135
3.4.2 Model Types . . . . . . . . . . . . . . . . . . . . . . 135
3.4.2.1 Cross-sectional models . . . . . . . . . . 136
3.4.2.2 Sector models . . . . . . . . . . . . . . . . 136
3.4.2.3 Cylindrical single well models . . . . . 137
3.4.2.4 Full field models . . . . . . . . . . . . . . 137
3.5 Grid Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
3.5.1 Cartesian Systems . . . . . . . . . . . . . . . . . . . 139
3.5.2 Cylindrical Systems . . . . . . . . . . . . . . . . . . 140
3.5.3 Stream Line Grids . . . . . . . . . . . . . . . . . . . 141
3.5.4 Special Connections . . . . . . . . . . . . . . . . . . 142
3.5.5 Corner Point Representation . . . . . . . . . . . . 144
3.5.6 Local Grid Refinement . . . . . . . . . . . . . . . . 144
3.5.7 Unstructured Grids . . . . . . . . . . . . . . . . . . 145
3.5.8 Surface Constraints . . . . . . . . . . . . . . . . . . 146
3.5.9 Non-Orthogonality . . . . . . . . . . . . . . . . . . 147
3.6 Rock Properties . . . . . . . . . . . . . . . . . . . . . . . . . 147
3.6.1 Core Data . . . . . . . . . . . . . . . . . . . . . . . . 148
3.6.1.1 Routine core analysis . . . . . . . . . . . 148
3.6.1.2 Geocellular up-scaling . . . . . . . . . . 150
3.6.1.3 Special core analysis . . . . . . . . . . . 150
3.6.2 Log Data . . . . . . . . . . . . . . . . . . . . . . . . . 157
3.6.3 Test Data . . . . . . . . . . . . . . . . . . . . . . . . 160
3.7 Model Relative Permeability . . . . . . . . . . . . . . . . . 161
August 11, 2017 10:20 Topics in Reservoir Management - 9in x 6in b2764-fm page xiii
Contents xiii
3.7.1 Data Manipulation . . . . . . . . . . . . . . . . . . 161

3.7.2 Three-Phase Relative Permeability . . . . . . . . 165
3.7.3 Vertical Equilibrium . . . . . . . . . . . . . . . . . 166
3.7.4 Pseudo Relative Permeability . . . . . . . . . . . . 168
3.7.5 Well Pseudo Relative Permeability . . . . . . . . 173
3.7.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . 174
3.8 Model Capillary Pressure . . . . . . . . . . . . . . . . . . . 174
3.8.1 Manipulation of Capillary Pressure . . . . . . . . 176
3.8.2 Vertical Equilibrium . . . . . . . . . . . . . . . . . 178
3.8.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . 179
3.9 Fluid Properties and Experiments . . . . . . . . . . . . . . 179
3.9.1 Single Component Properties . . . . . . . . . . . . 180
3.9.2 Properties of Mixtures . . . . . . . . . . . . . . . . 180
3.9.3 Hydrocarbon Types . . . . . . . . . . . . . . . . . . 182
3.9.4 Definitions . . . . . . . . . . . . . . . . . . . . . . . . 183
3.9.5 Experiments . . . . . . . . . . . . . . . . . . . . . . . 186
3.9.5.1 Constant composition expansion . . . 186
3.9.5.2 Differential liberation . . . . . . . . . . . 186
3.9.5.3 Constant volume depletion . . . . . . . 190
3.9.5.4 Separator tests . . . . . . . . . . . . . . . 190
3.10 Model Fluid Properties . . . . . . . . . . . . . . . . . . . . . 190
3.10.1 Black Oil Fluid Properties . . . . . . . . . . . . . . 190
3.10.2 Data Manipulation . . . . . . . . . . . . . . . . . . 193
3.10.3 Spatial Variations . . . . . . . . . . . . . . . . . . . 196
3.10.3.1 Variable bubble point . . . . . . . . . . . 197
3.10.3.2 Variable Api gravity . . . . . . . . . . . 197
3.11 Aquifer Treatment . . . . . . . . . . . . . . . . . . . . . . . . 197
3.11.1 Hurst Van Everdingen . . . . . . . . . . . . . . . . 198
3.11.2 Carter Tracy . . . . . . . . . . . . . . . . . . . . . . 199
3.11.3 Fetkovitch . . . . . . . . . . . . . . . . . . . . . . . . 199
3.11.4 Numerical Aquifer . . . . . . . . . . . . . . . . . . . 200
3.12 Model Well and Production Data . . . . . . . . . . . . . . 201
3.12.1 Well Inflow . . . . . . . . . . . . . . . . . . . . . . . 201
3.12.2 Production Control Data . . . . . . . . . . . . . . 203
3.12.2.1 Targets . . . . . . . . . . . . . . . . . . . . 203
August 11, 2017 10:20 Topics in Reservoir Management - 9in x 6in b2764-fm page xiv
xiv Topics in Reservoir Management
3.12.2.2 Constraints . . . . . . . . . . . . . . . . . 204

3.12.2.3 Actions . . . . . . . . . . . . . . . . . . . . 205
3.12.3 Practical Considerations . . . . . . . . . . . . . . . 205
3.12.3.1 History match . . . . . . . . . . . . . . . 205
3.12.3.2 Prediction . . . . . . . . . . . . . . . . . . 206
Chapter 4. History Matching 209

Deryck Bond
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209

4.2 Context of the History Match Study . . . . . . . . . . . . 211
4.2.1 The Business Context of the Study . . . . . . . . 211
4.2.2 Relation to Reservoir Development/
Management . . . . . . . . . . . . . . . . . . . . . . 212
4.2.3 The Work Flow Context of the Study . . . . . . 213
4.3 Static and Dynamic Data/Static and Dynamic
Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
4.3.1 Static and Dynamic Data . . . . . . . . . . . . . . 215
4.3.2 Dynamic Data and the Static Model . . . . . . . 216
4.4 Issues Related to Reservoir Simulation/Up-Scaling . . . 217
4.4.1 Issues Related to Grid Size/Numerical
Resolution . . . . . . . . . . . . . . . . . . . . . . . . 217
4.4.2 Issues Related to Representation Rate Variation
With Time . . . . . . . . . . . . . . . . . . . . . . . . 217
4.4.3 Issues Related to Representation of Well In-Flow
in the Simulator . . . . . . . . . . . . . . . . . . . . 218
4.5 Details of a “Conventional” Deterministic History
Match Study . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
4.5.1 Preparatory Work — Definition of Aims . . . . 219
4.5.2 Data Review/Well Histories . . . . . . . . . . . . . 219
4.5.3 Preparatory Work — Data Acquisition
Opportunities . . . . . . . . . . . . . . . . . . . . . . 221
4.5.4 Preparatory Work —“Classical” Reservoir
Engineering Calculations/Part Field Models . . 221
4.5.5 Review of the Simulation Model
(and Geological Model) . . . . . . . . . . . . . . . 222
August 11, 2017 10:20 Topics in Reservoir Management - 9in x 6in b2764-fm page xv
Contents xv
4.5.6
Outline of Approach to Matching the Model . . 223
4.5.7
How Well Should We Aim to Match Data? . . . 224
4.5.8
Assessing the “Goodness” of a Match . . . . . . 226
4.5.9
Well Controls During the History Match . . . . 227
4.5.10
Initial Simulation Runs . . . . . . . . . . . . . . . . 229
4.5.11
Review of Scope for Changing Model Input
Parameters . . . . . . . . . . . . . . . . . . . . . . . 230
4.5.12 Sensitivity Study . . . . . . . . . . . . . . . . . . . . 231
4.5.13 The History Matching Process . . . . . . . . . . . 232
4.5.14 Matching Pressure . . . . . . . . . . . . . . . . . . . 233
4.5.14.1 Pressure match example . . . . . . . . . 234
4.5.15 Matching Fluid Movement . . . . . . . . . . . . . 236
4.5.16 Matching Water Movement — Example 1 . . . 238
4.5.19 Matching Well Pressures/Detailed Well
Performance . . . . . . . . . . . . . . . . . . . . . .. 242
4.5.20 The Transition to Prediction . . . . . . . . . . .. 243
4.6 Automatic and Computer-Assisted History
Matching/Multiple Matches . . . . . . . . . . . . . . . . .. 243
4.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . .. 243
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 244
Index 245
August 11, 2017 10:19 Topics in Reservoir Management - 9in x 6in b2764-ch01 page 1
Chapter 1
Introduction to Rock Properties
Robert W. Zimmerman
Department of Earth Science and Engineering,
Imperial College London, Kensington, London SW7 2AZ, UK
1.1. Introduction
The most important fact about reservoir rocks is that, by definition,
they are not completely solid, but rather are porous to one degree
or another. The degree to which they are porous is quantified by a
parameter known as the porosity. The fact that the rocks are porous
allows them to hold fluid. If these pores are interconnected, which
they are in most rocks, then the fluid is able to flow through the
rock, and the rock is said to be permeable. The ability of a rock to
allow fluid to flow through it is quantified by a parameter called the
permeability. As the porosity controls the amount of oil or gas that
the rock can hold, and the permeability controls the rate at which
this oil or gas can flow to a well, these two parameters, porosity
and permeability, are the most important attributes of a rock, for
reservoir engineering purposes.
It would be very advantageous to petroleum engineers if the pore
space of a reservoir were completely filled with hydrocarbon fluid.
Unfortunately, this is never the case, and the pores always contain
a mixture of hydrocarbons and water. The relative amounts of oil,
gas or water are quantified in terms of parameters known as the fluid
saturations. These saturations are in turn controlled by the surface
interactions between the rock and the various fluids, which can be
1
2 Topics in Reservoir Management
described and quantified in terms of properties known as wettability

and surface tension.
The ability of a rock to store fluid, and the relationship between
the amount of fluid stored in the rock and the pressure of the fluid, is
related to the porosity, and specifically to the way that the porosity
changes as the pore pressure changes. The relationship between
porosity and pressure is quantified by an important mechanical
property known as the pore compressibility.
Aside from properties such as porosity, permeability and com-
pressibility, which are obviously of crucial importance in reservoir
engineering, there are other petrophysical (“petros” = rock, in
Greek) properties that are important, but for less obvious reasons.
One such property is the electrical resistivity. Although the resistivity
is not directly related to the oil, its value is controlled mainly by
the amount of water in the rock, and so knowledge of the electrical
resistivity gives us valuable information on the relative amounts of
oil and water in the rock.
In this introductory chapter on rock properties, we will define the
various parameters mentioned above, present some simple models
to relate these properties to the pore structure of the rock, and
give some indication of how these properties are used in petroleum
engineering.
1.2. Porosity and Saturation

1.2.1. Definition of Porosity
If one looks at a typical cylindrical core of rock, with radius R
and length L, it would have an apparent volume, or bulk volume, of
Vb = πR2 L. However, on a smaller scale, such as under a microscope
(Fig. 1.2.1), it would be clear that some of this volume is occupied
by rock minerals, and some of it is void space.
We can now define the mineral volume of this core, Vm , as the
volume that is actually occupied by minerals. Lastly, we define the
pore volume, Vp , as the volume of the void space contained in this
cylindrical core. These volumes are related by
Vb = Vm + Vp . (1.2.1)
Introduction to Rock Properties 3
Sand grains
Pore
Figure 1.2.1. Schematic diagram of porous sandstone, showing grains and pore
space. Typical grain sizes are tens to hundreds of microns.
The fraction of the cylinder that is occupied by pore space

is known as the porosity, and is usually denoted by φ (although
sometimes by n):
φ = Vp /Vb . (1.2.2)
The porosity of a reservoir rock can range from a few percent, to

as high as 40%.
A distinction is often made between primary porosity, which
is the porosity that the sandstone (say) had after it was first
deposited and compacted, and secondary porosity, which is any
porosity that is subsequently created through mineral dissolution,
mineral deposition, fracturing, etc. One very important type of
secondary porosity is the porosity contained in natural fractures.
Many reservoirs, which collectively contain roughly half of all known
oil reserves, are naturally fractured. These reservoirs are often filled
with an interconnected system of fractures. The porosity contained
in these fractures is usually in the order of 0.1–1.0%, and is much
less than the primary porosity, but the fracture network typically
has a much higher permeability than the unfractured rock, often by
several orders of magnitude. These types of reservoirs are referred
to as dual-porosity reservoirs. Producing oil from such reservoirs is
more problematic than producing oil from unfractured reservoirs.
Another distinction that can be made is between total porosity,
which is essentially the porosity that is defined by Eq. (1.2.2),
and effective porosity, which measures only the pore space that is
interconnected and which can potentially form a flow path for the
hydrocarbons. The total porosity is therefore composed of effec-
tive/interconnected porosity, and ineffective/unconnected porosity.
In most rocks, there is little ineffective porosity. One important
exception are carbonate rocks called diatomites, in which most
of the porosity is unconnected. The Belridge oilfield in central
California has produced 1.5 billion barrels of oil from a diatomite
reservoir that has porosities ranging from 45% to 75%, most of
which is unconnected and not “effective”! But this is a special case
that requires special production methods, and is not typical of the
reservoirs that will be the focus of most of this course.
1.2.2. Heterogeneity and “Representative

Elementary Volume”
The property of porosity introduces an issue, that of heterogeneity,
which is important for all petrophysical properties. To say that a
reservoir is “heterogeneous” means that its properties vary from
point-to-point. In one sense all rock masses are heterogeneous,
because, as we move away from a given rock at a given location, we
will eventually encounter a different rock type. For example, some
reservoirs contain beds of sands and shales, with thicknesses in the
order of a few metres, such as in Fig. 1.2.2. These reservoirs are
heterogeneous on length scales in the order of tens of metres.
Sand
10 m
Shale
Figure 1.2.2. Sand–shale sequence in a reservoir. Such a reservoir is homogeneous

on a length scale of a few centimetres, but heterogeneous on a length scale of a
few metres.
x1
x2
(a) (b)
Figure 1.2.3. (a) Porosity at x1 is 0, and at x2 is 1, illustrating that porosity

must not be defined at a point, but over a volume. (b) φ(R) can be defined as an
average over a region of radius R (length of the arrow).
At the other extreme of length-scale, all porous rocks are

heterogeneous on the scale of the pore size. Consider Fig. 1.2.3(a),
where x1 and x2 are two locations in the rock. If we ask “what is
the porosity at location x1 ”?, then the answer would be “0”, because
point x1 lies in a sand grain. On the other hand, point x2 lies in a
pore, so, strictly speaking, the porosity at x2 is 1. Clearly, it makes no
sense to talk about the porosity at some infinitely small mathematical
point in the reservoir. When we discuss the porosity, we implicitly
are referring to the average porosity in some small region.
Imagine that we could measure the porosity in a small spher-
ical region of rock, of radius R, surrounding the point x2 , as in
Fig. 1.2.3(b). Let’s denote this average value as φ(R). For very small
values of R, φ(R) would be 1. As R gets larger, the spherical region
will start to encompass some of the nearby sand grains, and so φ(R)
will decrease. In a typical sandstone, φ(R) will fluctuate with R, but
eventually stabilise to some constant value. Eventually, as R gets
large enough that the region crosses over into the next rock type,
φ(R) might change abruptly. Schematically, we can represent this
situation as in Fig. 1.2.4 (see Bear, 1972).
The minimum value of R needed for the porosity to stabilise
is known as the “representative elementary volume” (REV), or as
the “representative volume element” (RVE). When we talk about
0 R
REV
Figure 1.2.4. Porosity as a function of the size of the sampling region, showing
the existence of an REV.
the porosity, we are usually implicitly referring to the porosity as

defined on a length scale at least as large as the REV.
For sandstones with a uniform grain size distribution, the
REV must be at least about ten grain diameters. However, for
heterogeneous rocks such as many carbonates, the REV may be much
larger, as heterogeneity may exist at many scales. In fact, there is no
guarantee that a REV scale exists for a given rock. But in petroleum
engineering, we always assume that a REV can be defined on the
length scale of the gridblocks that are used in the numerical reservoir
simulation codes. This will be discussed further in some later sections.
1.2.3. Saturation
Reservoir rocks are never filled completely with oil, for reasons that
will be discussed later in this chapter. Consider a rock that contains
some oil and some water, as in Fig. 1.2.5. If the volume of water
contained in a region of rock is Vw , the volume of oil is Vo , and the
total volume of the pore space is Vp , then the saturation of each
of these phases can be defined as the fraction of pore space that is
occupied by that phase, i.e.
Sw = Vw /Vp , So = Vo /Vp . (1.2.3)
The saturation of each phase must, by definition, lie between 0 and 1.

If oil and water are the only two phases present, then it is necessarily
Sand
Oil grain
Water Pore
Figure 1.2.5. Schematic diagram of a porous rock containing oil and water.
the case that,
Sw + So = 1. (1.2.4)
If there is also some hydrocarbon gas in the pore space, then
Sw + So + Sg = 1. (1.2.5)
1.3. Permeability and Darcy’s Law

1.3.1. Darcy’s Law
The ability of a porous rock to transmit fluid is quantified by the
property called permeability. Quantitatively, permeability is defined
by the “law” that governs the flow of fluids through porous media —
Darcy’s law. This law was formulated by French civil engineer Henry
Darcy in 1856 on the basis of his experiments on water filtration
through sand beds. Darcy’s law is the most important equation in
petroleum engineering.
Imagine a fluid having viscosity µ, flowing through a horizontal
tube of length L and cross-sectional area A, filled with a rock or
sand. The fluid pressure at the inlet is Pi , and at the outlet is Po , as
in Fig. 1.3.1.
According to Darcy’s law, the fluid will flow through the rock
in the direction from higher pressure to lower pressure, and the
volumetric flowrate of this fluid will be given by
kA(Pi − Po )
Q= , (1.3.1)
µL
Pi
Po
L
x
Q
A
Figure 1.3.1. Experimental setup for measuring the permeability of a porous rock
or sand.
where
Q= volumetric flowrate, with units of m3 /s,

k= permeability of the rock, with units of m2 ,
A= cross-sectional area of the rock core, with units of m2 ,
Pi , Po = inlet/outlet pressures, with units of Pa,
µ= viscosity of the fluid, with units of Pa·s,
L= length of the core, with units of m.
This equation can be thought of as providing a definition of

permeability, and it also shows us how to measure the permeability
in the laboratory. This equation tells us that the flowrate is pro-
portional to the area, inversely proportional to the fluid viscosity,
and proportional to the pressure gradient, i.e. the pressure drop per
unit length, ∆P /L. Note that the permeability is a property of the
rock; the influence of the fluid that is flowing through the rock is
represented by the viscosity term in Darcy’s law.
It is usually more convenient to work with the volumetric flow
per unit area, q = Q/A. Darcy’s law is therefore usually written as
Q k(Pi − Po )
q= = , (1.3.2)
A µL
where the flux q has dimensions of m/s. Please note that the flux
is not the same as the velocity of the fluid particles,1 and so it is
perhaps easier to think of these units as m3 /m2 ·s.
For the general case in which the flux may vary from point-
to-point, we need a differential form of Darcy’s law. The differential
version of Eq. (1.3.2) for horizontal flow is,
−k dP
qx = . (1.3.3)
µ dx
The minus sign is included to account for the fact that the fluid flows
in the direction from higher to lower pressure.
For vertical flow, we must include a gravitational term in Darcy’s
law. To see why this is necessary, recall from fluid mechanics that if
the fluid is stagnant, then the pressure distribution will be
P = Po + ρgz, (1.3.4)
where z is the depth below some datum level, and Po is the pressure
at the datum level. So, there will be pressure gradient in a stagnant
fluid, but there will be no flow. The “equilibrium pressure gradient”
is, from Eq. (1.3.4),

dP
= ρg. (1.3.5)
dz equilibrium
It seems reasonable to assume that fluid will flow through the

rock only if the pressure gradient exceeds the equilibrium value
given by Eq. (1.3.5). In this case, the actual driving force should
1
Note: q is the flux based on the total nominal area of the core, but the fluid
actually flows only through the pores, and not the grains! So, the total flux is
given by Q = qA, but it can also be expressed as Q = vApore , where v is the
actual mean velocity of the particles of fluid, and Apore is area occupied by pores.
Hence, qA = vApore , so v = q(A/Apore ) = q/φ. For example, if q is 1 cm/h in a
reservoir of 10% porosity, the actual mean velocity of the oil molecules as they
travel through the pore space is 10 cm/h.
be (dP/dz) − ρg. For vertical flow, we therefore modify Eq. (1.3.3) as

follows:

−k dP −k d(P − ρgz)
qz = − ρg = . (1.3.6)
µ dz µ dz
Actually, this form of the equation holds for horizontal flow also,
because in this case we can say that
−k d(P − ρgz) −k dP
qx = = , (1.3.7)
µ dx µ dx
since d(ρgz)/dx = 0.
A convenient way of simplifying the form of these equations is to
write them in terms of the fluid potential Φ, defined by
Φ = P − ρgz, (1.3.8)
in which case flow in an arbitrary direction n can be described by2

−k dΦ
qn = . (1.3.9)
µ dn
The above equations assume that the permeability is the same
in all directions, but in most reservoirs the permeability in the
horizontal plane, kH , is different than the vertical permeability, kV ;
typically, kH > kV . The permeabilities in different directions within
the horizontal plane may also differ, but this difference is usually not
as great as that between kH and kV . The property of having different
permeabilities in different directions is known as anisotropy. For flow
in an anisotropic rock, we must modify Darcy’s law, as follows:
−kH dΦ −kv dΦ
qx = , qz = . (1.3.10)
µ dx µ dz
2
Caution: if the rock is anisotropic, then Eq. (1.3.9) does not hold in an arbitrary
direction, even if we use an appropriate value of k (de Marsily, 1986). The correct
version of Darcy’s law for an anisotropic rock must be written in terms of the
permeability tensor, which is a symmetric 3 × 3 matrix that has six independent
components. However, this tensorial form of Darcy’s law is not typically used in
most reservoir engineering calculations.
Another way to think about why fluid flow is controlled by the

gradient of P − ρgz is as follows. You may recall from undergraduate
fluid mechanics that Bernoulli’s equation, which essentially embodies
the principle of “conservation of energy”, contains the terms

P v2 1 ρv 2
− gz + = P − ρgz + , (1.3.11)
ρ 2 ρ 2
where P/ρ is related to the enthalpy per unit mass, gz is the

gravitational potential energy per unit mass, and v 2 /2 is the kinetic
energy per unit mass.
Fluid velocities in a reservoir are usually very small, so the kinetic
energy term is negligible, in which case the combination P − ρgz
represents the “Bernoulli energy”. It seems reasonable that the fluid
would flow from regions of higher to lower energy, and, therefore, the
driving force for flow should be the gradient (i.e. the rate of spatial
change) of the quantity P − ρgz.
These considerations also warn us that we should not expect
Darcy’s law to hold in cases where the kinetic energy term is not
negligible. In fact, at high flowrates, we must modify Darcy’s law
by incorporating a quadratic term q 2 on the left-hand side of,
say, Eq. (1.3.3). The resulting more general equation, called the
Forchheimer equation, is necessary in some situations, such as in
some gas reservoirs, and particularly near the wellbore, where the
velocities are higher (Bear, 1972). However, Darcy’s law is adequate
in the vast majority of situations.
1.3.2. Units of Permeability

Permeability has dimensions of “area”, so in the SI system it has
units of m2 . However, in most areas of engineering it is conventional
to use a unit called the “Darcy”, which is defined by
1 Darcy = 0.987 × 10−12 m2 ≈ 10−12 m2 . (1.3.12)
The Darcy unit is defined such that a rock having a permeability

of 1 Darcy would transmit 1 cm3 of water (which has a viscosity of
Table 1.3.1. Typical ranges of the permeability

of various rocks and sands.
Rock type k (D) k (m2 )
Coarse gravel 103 –104 10−9 –10−8

Sands gravels 100 –103 10−12 –10−9
Fine sand, silt 10−4 –100 10−16 –10−12
Clay, shales 10−9 –10−6 10−21 –10−18
Limestones 10−4 –100 10−16 –10−12
Sandstones 10−5 –101 10−17 –10−11
Weathered chalk 100 –102 10−12 –10−10
Unweathered chalk 10−9 –10−1 10−21 –10−13
Granite, gneiss 10−8 –10−4 10−20 –10−16
1 centipoise/s) through a region that has a cross-sectional area of

1 cm2 , if the pressure drop along the direction of flow was 1 atm/cm.
This definition is strange, in that it utilises different systems of
units. Some people apply Darcy’s law by first converting flowrates
to cm3 /s, converting areas to cm2 , etc., in which case you must use
the value of k in Darcies. Another method is to first convert all
parameters to SI units, in which case you must then use the value of
k in units of m2 when you apply Darcy’s law.
Typical ranges of the permeability of various rocks and sands are
given in Table 1.3.1.
Table 1.3.1 shows that the permeability of geological media varies
over many orders of magnitude. However, most reservoir rocks have
permeabilities in the range of 0.1 mD to 10 D, and usually in the much
narrower range of 10–1000 mD. Methods for measuring permeability
in the laboratory are discussed in the chapter on core analysis.
1.3.3. Relationship between Permeability

and Pore Size
The permeability depends on the porosity of the rock, and also on
the pore size. Many models have been developed to try to relate the
permeability to porosity, pore size, and other attributes of the pore
space. The simplest model assumes that the pores are all circular
tubes of the same diameter. Consider a set of circular pore tubes,
d L
Figure 1.3.2. Idealised pore structure used to derive a relationship between

permeability, porosity and pore size.
each of diameter d, passing through a cubical rock specimen of side

L, as in Fig. 1.3.2, with a pressure difference ∆P imposed across the
two parallel faces (on the page and behind the page) of the rock.
According to Poiseuille’s equation for pipe flow (Dullien, 1992),
the flow through each tube is given by
π d4 ∆P
Q= . (1.3.13)
128µ L
If there are N such pores, the total flowrate will be
N π d4 ∆P
Q= . (1.3.14)
128µ L
The total area of these pores, in the plane of the page, is Ap =
N π d2 /4, and the porosity is φ = Ap /A = Ap /L2 , where A is the
macroscopic area normal to the flow. Hence, the total flowrate from
Eq. (1.3.14) can be written as
φd2 A ∆P
Q= . (1.3.15)
32µ L
If we compare this flowrate with Darcy’s law, Q = kA∆P/µL,

we see that the permeability of this rock is k = φd2 /32. Lastly, we
note that an isotropic rock should have only one third of its pores
aligned in the x-direction, one third in the y-direction, etc. So, the
permeability of this idealised porous rock is
φd2
k= . (1.3.16)
96
A slightly more realistic model, in which the orientations of the pores

are randomly distributed in 3D space, leads to exactly the same result
(Scheidegger, 1974).
Equation (1.3.16) is often written in terms of the “specific
surface”, S/V , which is the total amount of surface area (S) per
unit volume of rock; the result is
φ3
k= . (1.3.17)
6(S/V )2
This is often called the Kozeny–Carman equation (Bear, 1972). One
justification for this equation is that the permeability is the inverse
of the “hydraulic resistivity”, and it is plausible that the resistance
to flow, which is essentially due to viscous drag of the fluid against
the pore walls, would be related to the amount of surface area of the
pores.
In some versions of the Kozeny–Carman equation, the numerical
factor six is replaced by another constant called the “tortuosity”, τ .
The tortuosity is sometimes claimed to represent the ratio of the
actual fluid flow path from the inlet to the outlet, to the nominal
fluid flow path L, but this is not true, and it is preferable to think
of it as nothing more than an empirical fitting factor.
There have been many attempts to try to improve upon the
Kozeny–Carman equation, by incorporating more information about
the distribution of pore sizes, interconnectedness of the pores, etc.
For the purposes of this chapter, it is sufficient to understand that the
permeability is proportional to the square of the mean pore diameter.
1.3.4. Permeability of Layered Rocks

Most reservoir rocks are layered, with each layer having a different
permeability. If fluid flows through a layered rock, either in the
vertical direction (perpendicular to the layering) or the horizontal
direction (parallel to the layering), it is possible to define an effective
permeability that will allow us to treat the rock as if it were
homogeneous, and use Darcy’s law in its usual form.
For example, consider horizontal flow through a rock composed
of N layers, each having permeability ki and thickness Hi , as in
∆Φ
Q1 k1 H1
Q2 k2 H2
.....
QN kN HN
Figure 1.3.3. Fluid flow parallel to the layering of a layered rock.
Fig. 1.3.3. Within each layer, fluid will flow horizontally, according
to Darcy’s law:
−ki (Hi w) ∆Φ
Qi = , (1.3.18)
µ ∆x
where w is the thickness into the page.
The total flowrate is found by summing up the flowrates through
each layer:
N
N
N
−ki (Hi w) ∆Φ −w ∆Φ
Q= Qi = = ki Hi , (1.3.19)
µ ∆x µ ∆x
i=1 i=1 i=1
but if we were to treat the rock as if it were a homogeneous rock

mass with an effective permeability keff , then we would write Darcy’s
law as
N
−keff (Htotal w) ∆Φ −w ∆Φ
Q= = keff Hi . (1.3.20)
µ ∆x µ ∆x
i=1
If we compare Eqs. (1.3.19) and (1.3.20), we see that the effective

permeability of the layered rock is
N
N N
1
keff = ki Hi Hi = ki Hi . (1.3.21)
H
i=1 i=1 i=1
Hence, the effective permeability for flow along the layering is the
weighted arithmetic mean of the individual permeabilities, weighted
by the thickness of the layers.
Q
Area, A
k1 H1
k2 H2
∆Φ
.....
kN yN
Figure 1.3.4. Fluid flow perpendicular to the layering through a layered rock.
Now imagine vertical flow through this layered system

(Fig. 1.3.4). We again start by writing Darcy’s law for each layer:
−ki A ∆Φi
Qi = , (1.3.22)
µ Hi
where A is the area normal to the flow direction, i.e. in the horizontal
plane.
In the steady state, the flowrate through each layer must be the
same, but the potential drop will be different. So, we put Qi = Q in
each layer, and rewrite Eq. (1.3.22) in the form
−µQHi
∆Φi = . (1.3.23)
Aki
The total potential drop across all N layers is found by summing up

the drops across each individual layer:
N
N
N
−µQHi −µQ Hi
∆Φ = ∆Φi = = . (1.3.24)
Aki A ki
i=1 i=1 i=1
The potential drop across an “equivalent” homogeneous rock of

total vertical thickness Hand area A would be
N
−µQH −µQ
∆Φ = = Hi . (1.3.25)
Akeff Akeff
i=1
Comparison of Eqs. (1.3.24) and (1.3.25) shows that

N
N
N
−1
Hi 1 Hi
keff = Hi = . (1.3.26)
ki H ki
i=1 i=1 i=1
The expression on the right side of Eq. (1.3.26) is called the weighted
harmonic mean of the permeabilities.
Equations (1.3.21) and (1.3.26) are similar to the equations for
the overall conductivity of electrical resistors in parallel or series.
However, this analogy can easily be remembered incorrectly, because
the thickness also appears in these equations, and it appears in a
different way in the two cases. Rather than trying to remember the
analogy between electrical circuits and flow through layered rocks,
it is safer to derive the laws for the effective permeability from first
principles (or to refer to these notes).
Roughly speaking, the effective permeability for flow parallel to
the layering is controlled by the permeability of the most permeable
layer, whereas for flow transverse to the layering, the least permeable
layer plays the controlling role.
1.3.5. Permeability Heterogeneity

In many reservoirs the heterogeneity is more complex than the simple
layering shown in Figs. 1.3.3 and 1.3.4. Moreover, if oil is flowing
towards a well, the flow geometry will be radial, and clearly the
“series” and “parallel” models cannot be expected to apply.
In order to calculate fluid flow in a reservoir using either
analytical or numerical methods (both of which will be covered
later in this volume), it is necessary to replace the heterogeneous
distribution of permeabilities with a single “effective” permeability.
This difficult problem, which is known in petroleum engineering as
“upscaling”, will be covered in detail in a later volume.
For now, we only mention that the geometric mean often provides
a good estimate of the effective permeability, for cases in which the
heterogeneity is “randomly” distributed. The geometric mean of a
permeability distribution is defined such that the natural logarithm
of the geometric mean is the volumetrically-weighted average of the
logarithm of the individual permeabilities. For example, if we have

N different regions, “randomly” arranged, each with permeability ki
and volume fraction ci , then the geometric mean is defined as
N
N
N

ln(kG ) = ci ln(ki ) = ln(kici ) = ln kici , (1.3.27)
i=1 i=1 i=1
which is to say that

N
cN
kG = kici = kic1 kic2 kic3 · · · kN . (1.3.28)
i=1
It can be proven that the geometric mean always lies between

the arithmetic and harmonic means. Moreover, it can also be proven
that, regardless of the precise geometric distribution of the local
permeabilities, the effective permeability will always lie between the
arithmetic and harmonic mean values (Beran, 1968). The fact that
both the effective permeability and the geometric mean permeability
are bounded by the arithmetic and harmonic means provides some
justification for using the geometric mean as an approximation to
the effective permeability, in cases in which the heterogeneity is
“randomly” distributed.
1.4. Surface Tension, Wettability and Capillarity

The pore space of a reservoir rock always contains a mixture
of different fluids. The manner in which these fluids distribute
themselves within the pore space depends on the physico-chemical
interactions between the various fluids and rock minerals. We now
discuss some of the concepts and definitions needed to understand
the distribution of fluids in the pore space.
1.4.1. Surface Tension

Consider an interface between two fluids, which for concreteness we
take to be a gas and a liquid, as shown in Fig. 1.4.1. Now consider
a molecule within the liquid, such as molecule A on the right. This
molecule has a certain internal energy, u. There is a force of attraction
between this molecule and all adjacent liquid molecules; call it FLL .
Interface
FLL FLL
FGL B FLL FLL A FLL
FLL FLL
Gas Liquid
Figure 1.4.1. Simplified sketch of the forces acting on a fluid molecule.
The force of attraction between a liquid molecule and a gas molecule

will be called FGL .
Now imagine that we slowly pull molecule A towards the inter-
face. Initially, it will be attracted equally to all of its neighbouring
liquid molecules, and the net force on it will be zero. As it nears
the interface, the liquid molecule to its right will be pulling on it
with force FLL , while the gas molecule to its left will be pulling with
force FGL . For concreteness, assume that FGL < FLL . In this case,
there will be a net rightward force acting on molecule A, due to its
neighbouring molecules. If we want to pull A to the surface, we must
exert a leftwards force on it, and thereby do work on it. It follows
that when molecule A reaches the surface, it will have greater energy
than it did when it was in the bulk liquid.
The total excess energy that the liquid has due to its interface will
obviously be proportional to the number of molecules at the surface,
which is to say it will be proportional to the area of the interface.
We can modify the usual thermodynamic expression for the internal
energy to include surface energy as follows:
U = T S − P V + γA, (1.4.1)
where U is the internal energy, T is the temperature, S is the entropy,

P is the pressure, V is the volume, A is the interface area, and γ is
the “surface tension” between the liquid and the gas. If we treat the
extensive variables (S, V, A) as the independent variables, then the
Wire frame
L
b F
Loop
Figure 1.4.2. Thought experiment used to illustrate the significance of surface

tension.
differential of Eq. (1.4.1) is
dU = TdS − PdV + γdA. (1.4.2)
Imagine now that we have a thin film of liquid in a device such

as shown in Fig. 1.4.2. If we pull slowly on the slidable vertical bar
with a force F , and move this bar by a small distance dL, then the
work done by the external force on the liquid film will be FdL. By the
first law of thermodynamics, the work done must equal the change
in internal energy, so
dU = dW = FdL. (1.4.3)
If we pull on the bar slowly and adiabatically (i.e. “reversibly”),

the entropy change of the liquid film will be zero. Furthermore, the
volume of the film is negligible, so PdV will be essentially zero. So,
by Eq. (1.4.2),
dU = γdA. (1.4.4)
However, A = bL, and so dA = bdL, and Eq. (1.4.4) then gives
dU = γbdL. (1.4.5)
Equating Eqs. (1.4.3) and (1.4.5) shows that
FdL = γbdL ⇒ F = γb. (1.4.6)
In other words, the external effect of surface tension is the same as

if the interface were exerting a force of magnitude γ per unit length
of the edge of the interface. Because of this interpretation, it is often
convenient to treat an interface like an elastic membrane that exerts

a force along its perimeter. Equation (1.4.6) also implies that γ has
dimensions of force/length, and so it has SI units of N/m. Typical
values for an oil/water interface are 0.01–0.05 N/m.
1.4.2. Capillary Pressure

Because of surface tension, the pressures within two fluid phases
that are in mechanical equilibrium with each other across a curved
interface will not be equal. To prove this, consider a bubble of gas,
of radius R, inside a liquid that is contained in a rigid, thermally
insulated container, as shown in Fig. 1.4.3. The pressure in the gas
is PG , the pressure in the liquid is PL , and the surface tension of the
gas–liquid interface is γ.
According to Eq. (1.4.2), the total differential of the internal
energy of this liquid + gas + interface system will be
dU = TdS − PL dVL − PG dVG + γdA. (1.4.7)
Note, that we count the “volumetric” term for both the liquid
and the gas, but we must count the interface only once. However,
VG + VL = Vcontainer = constant, so dVL = −dVG ; hence
dU = TdS + PL dVG − PG dVG + γdA. (1.4.8)
Now assume that the bubble grows slowly. Since the container
is rigid and thermally insulated, dS = 0 and dU = 0 (i.e. no heat is
PL
PG 2R
Figure 1.4.3. Thought experiment of a gas bubble surrounded by liquid, encased

in a rigid, thermally insulated container, used to derive the Young–Laplace
equation.
added to the system, and no work is done on the system, so the total
entropy and energy remain the same). Hence,
γdA = PG dVG − PL dVG = (PG − PL )dVG . (1.4.9)
But A = 4πR2 , so dA = 8πRdR, and V = 4πR3 /3, so dV = 4πR2 dR.

Hence, Eq. (1.4.9) can be written as
8πγRdR = (PG − PL )4πR2 dR, (1.4.10)
which is equivalent to
PG − PL = 2γ/R. (1.4.11)
This is the famous Young–Laplace equation, which states that the

pressure inside the bubble is greater than the pressure outside,
by an amount that is directly proportional to the surface tension
between the two fluids, and inversely proportional to the radius of
the bubble.
This pressure difference is known as the capillary pressure, i.e.
PG − PL ≡ Pcap = 2γ/R. (1.4.12)
A capillary pressure difference exists when any two fluids are in

contact, not necessarily a liquid and a gas. For example, if we have
a bubble of oil surrounded by water, then Eq. (1.4.12) would hold,
with the subscripts G and L replaced by o for oil and w for water.
In a rock that is filled with oil and water, the interface between
these two phases will be locally curved, and the pressures in the water
and the oil phases will differ by an amount given by Eq. (1.4.12),
where we define R to be the mean radius of curvature of the interface.
Because of the inverse dependence of capillary pressure on radius,
capillary pressures are more important in rocks with smaller pores
than for rocks having larger pores.
1.4.3. Contact Angles

We now consider what happens when two fluids are in contact with
a solid surface, as in Fig. 1.4.4, where a drop of liquid is sitting on a
solid surface, surrounded by gas.
Gas
α Liquid
Solid
Figure 1.4.4. A drop of liquid sitting on a solid surface, surrounded by a gaseous

phase.
γLG
Gas Liquid
γGS α γLS
Solid
Figure 1.4.5. Force balance on the contact between three phases (liquid sitting
on a solid surface, surrounded by a gaseous phase).
The slanted line is the tangent to the gas–liquid interface at the

point where the interface meets the solid surface. The angle between
the solid surface and the tangent, measured by rotating the solid
surface towards the tangent, passing through the liquid (by definition
the angle is measured through the denser phase), is called the contact
angle, α.
We now do a free-body diagram (Fig. 1.4.5) and force-balance
on the region where the three phases (solid, liquid, gas) meet; this
is exactly like the “method of joints” used in analysing structural
trusses. When we “slice” through each interface, the part of the
interface that is “removed” will exert a tension γH on the “joint”,
where H is the distance into the page. For each surface tension, we
will use subscripts to denote the two fluids that form the interface; i.e.
γLS is the surface tension of the liquid–solid interface, etc. A force-
balance in the horizontal direction yields
ΣFhorizontal = γLS − γGS + γLG cos α = 0,

⇒ cos α = (γGS − γLS )/γLG . (1.4.13)
Several different cases can arise, depending on the relative

magnitudes of the three surface tensions.
Case 1: 0 < γGS − γLS < γLG :

In this case, the interfacial energy of a gas–solid interface is greater
than that of a liquid–solid interface, so, roughly speaking, the solid
will “prefer” to be in contact with the liquid. The right-hand side of
Eq. (1.4.13) will lie between 0 and 1, so α will lie in the range
0 < α < 90◦ . (1.4.14)
In this case, we say that the liquid “wets” the solid surface, and
the surface is called “water-wet” (although a better name would
be “water-wettable”, since a “water-wet” surface can be completely
dry!).
Case 2: γLG < γGS − γLS < 0:
In this case, the interfacial energy of a gas–solid interface is less than
that of the liquid–solid interface, so the solid will “prefer” to be in
contact with the gas. The right-hand side of Eq. (1.4.13) will lie
between −1 and 0, and so α will lie in the range
90◦ < α < 180◦ . (1.4.15)
In this case, we say that the liquid does not wet the surface; it
will sit on the surface in a bubble-shape, with very little interfacial
contact between the liquid and the solid surface; see Fig. 1.4.6.
Case 3: |γGS − γLS | > γLG :
In this case, the right-hand side of Eq. (1.4.13) does not lie between
−1 and +1, and so there is no value of α that will satisfy the equation
of mechanical equilibrium! To see what will happen in this case,
consider the limiting case in which γGS − γLS = γLG , in which case
“Wetting” liquid “Non-wetting” liquid
Gas α
α
Solid
Figure 1.4.6. A wetting liquid (left) and a non-wetting liquid (right) on a solid
surface.
cosα = 1, and α = 0◦ . In this case, the liquid will spread out over
the surface, creating as much liquid–solid interfacial area as possible
(example: oil on water!). If γGS −γLS > γLG , then the same situation
will occur, and the liquid will continue to flow until it forms a thin
layer on the solid surface. By a similar argument, if γGS − γLS <
−γLG , then the gas will spread out to cover as much of the solid
surface as possible.
Most reservoir rocks are preferentially “water-wet” as opposed
to “oil-wet”. If a water-wet rock is partially saturated with oil and
water, the pore walls will “prefer” to be in contact with water rather
than with oil, and so the oil will tend to exist in the form of blobs,
as in Fig. 1.4.6.
1.4.4. Capillary Rise

Consider a bucket containing some oil and some water, as in
Fig. 1.4.7(a). These two fluids are not miscible, and oil is usually
less dense than water, so the oil will sit on top of the water. The air
on top of the oil is at atmospheric pressure.
Now imagine that we put a capillary tube, of radius R and
composed of a water-wet material, in the bucket, as in Fig. 1.4.7(b).
This tube can be thought of as a simple model of a porous rock. The
T
Air Air θ P
A
D z
Oil Oil
Water Water
C P
(a) (b) (c)
Figure 1.4.7. (a) Oil, air and water in a bucket; oil lighter than water. (b) If a
capillary tube is inserted into the water, the water rises in the tube to a higher
level. (c) Force balance on the column of water in the tube.
water will rise up in the tube by some height, h, above the original
oil–water contact.
We now calculate the capillary rise, h, by performing a vertical
force balance on the column of water in the tube, as in Fig. 1.4.7(c).
The bottom of the column is pushed upwards by the pressure in the
water at level C, acting over an area of πR2 . With the sign convention
that the z-axis decreases with depth, this force is −Pw (zC )πR2 .
At the top of the column is a downwards-acting force due to the
pressure in the oil at level zD ; this force is +Po (zD )πR2 . The surface
tension exerts an upwards force along the entire wetted perimeter of
the tube; its magnitude is T = 2πγow R. It acts at an angle θ to the
vertical, so its vertical component is −2πγow R cos θ. Finally, gravity
acts downwards on the column of water with a force
W = mg = ρw Vg = ρw πR2 (zC − zD )g. (1.4.16)
Summing all the vertical forces to zero gives
−Pw (zC )πR2 + Po (zD )πR2 − 2πγR cos θ + ρw πR2 (zC − zD )g = 0.
(1.4.17)
The pressure in the oil at location D is equal to atmospheric
pressure plus the pressure due to a column of oil of height (zD − zA ),
i.e. Po (zD ) = Patm +ρo g(zD −zA ). Similarly, we can see that Pw (zC ) =
Patm + ρo g(zB − zA ) +ρw g(zC − zB ). Inserting these expressions into
Eq. (1.4.17) gives
−[Patm + ρo g(zB − zA ) + ρw g(zC − zB )]πR2
+ [Patm + ρo g(zD − zA )]πR2
− 2πγow R cos θ + ρw πR2 (zC − zD )g = 0, (1.4.18)
which can be solved to find the capillary rise, h:
2γow cos θ
zB − zD = h = . (1.4.19)
(ρw − ρo )gR
Hence, the height to which water would rise in a tube of radius R is
proportional to the surface tension between the water and oil, and is
inversely proportional to the radius of the tube.
We now examine the difference in pressures between the oil at

depth D (outside the tube), and the water at depth D (inside the
tube). By definition, this is the capillary pressure at depth D.
First, recall that Po (zD ) = Patm +ρo g(zD −zA ). Next, by starting
at point A, going down to point B in the oil, and then going back up
to point D in the water, we can find that Pw (zD ) = Patm + ρo g(zB −
zA ) − ρw g(zB − zD ). Hence,
Po (zD ) − Pw (zD )
= Patm + ρo g(zD − zA ) − Patm − ρo g(zB − zA ) + ρw g(zB − zD )
= (ρw − ρo )g(zB − zD ) = (ρw − ρo )gh = 2γow cos θ/R. (1.4.20)
In other words, Pcap = 2γow cos θ/R. This is identical to the Young–
Laplace equation that we derived earlier for an oil bubble in water,
modified to account for the contact angle. Note also that the capillary
pressure at any height h is equal to (ρw − ρo )gh. The theory we have
just described is referred to as “capillary-gravity equilibrium”.
1.4.5. Oil–Water Transition Zone

Most oils are less dense than water, so we might expect that a
reservoir would contain only oil down to a certain depth, and only
water below that depth, as would occur in a bucket containing oil
and water. Although it is true that the rock is usually fully saturated
with water below a certain level, on top of this zone is an oil–water
transition zone, in which the water saturation decreases gradually
with height. We can use the concept of capillary rise in a tube, along
with a modified form of the parallel-tube model of a porous medium,
to understand the existence of this oil–water transition zone in a
reservoir.
We first return to our parallel tube model of a porous rock, but
now imagine a distribution of different radii. Now, imagine that we
place this porous rock into our bucket of oil and water:
According to Eq. (1.4.19), water would rise very slightly into a
pore that has a large radius, but would rise very high in a small pore.
If we have a distribution of pore radii, then at an elevation such as
A in Fig. 1.4.8, all of the pores would be filled with water, and the
B
Oil
Water
Figure 1.4.8. Similar to Fig. 1.16(b), but with a set of capillary tubes of different
radii. According to Eq. (1.4.19), the water will rise higher in the smaller tubes.
Capillary Height above

pressure, gh the free water
(different scale) level, h
Oil/water contact
Free water level
0 Swi 1
Sw
Figure 1.4.9. Capillary pressure (left scale) and height above the FWL (right
scale), as functions of water saturation.
water saturation would be Sw = 1. At elevation B, some of the pores

will be filled with water, and others with oil, so the water saturation
will be 0 < Sw < 1. Finally, at a high enough elevation, such as C,
all the pores will be filled with oil, and Sw = 0.
Hence, the water saturation will be a decreasing function of the
height h above the “free water level” (FWL), which is defined as the
highest point in the reservoir where the capillary pressure is zero; see
Fig. 1.4.9. According to our capillary tube model shown in Fig. 1.4.8,
the water saturation will actually be equal to 1 up to some finite
height above the FWL that is controlled by the radius of the largest
pores. The highest point at which the saturation is equal to 1 is
known as the “oil–water contact” (OWC). Note that in a water-wet
reservoir, the OWC is above the FWL.
From Eq. (1.4.20), the “capillary pressure” is given by
Pcap = Po (h) − Pw (h) = (ρw − ρo )gh, (1.4.21)
and so the y-axis in Fig. 1.4.9 essentially represents both the capillary
pressure and the height above the FWL, which differ only by the
multiplicative factor (ρw − ρo )g. Hence, this graph represents the
water–oil transition zone, but also represents the capillary pressure
function as a function of saturation.
Note that the relation Pcap = (ρw − ρo )gh holds regardless of
the specific rock geometry, assuming only that the rock is water-wet.
However, the precise form of the Pcap (Sw ) curve shown in Fig. 1.4.9
depends on the pore geometry, and specifically on the pore-size
distribution.
For the simple bundle-of-parallel-tubes model, one can derive
an exact relationship between the Pcap (Sw ) curve and the pore-size
distribution. For a real rock, in which the pores are interconnected,
the relationship is not so simple, but it is always true that a
narrow pore-size distribution corresponds to a Pc curve with a nearly
horizontal shape, whereas a broader pore-size distribution yields a
curve that increases more gradually, as shown in Fig. 1.4.10. The
extreme case of a bundle of tubes in which all pores had the same
Broad pore-size
distribution
Pcap
Narrow pore-size
distribution
0 Swi Sw 1
Figure 1.4.10. Capillary pressure curves for two rocks, with a narrow and a broad
pore-size distribution, respectively.
radius R would yield a capillary pressure function that was essentially

a horizontal line, at the value Pcap = 2γow cos θ/R.
Note that whereas the bundle-of-tubes model predicts that the
water saturation becomes zero at some (large) height above the
FWL, in a real rock the water saturation never falls below some
non-zero value Swi , known as the irreducible water saturation, which
is typically about 10% (Fig. 1.4.10). Hence, there is generally no
region in a water-wet reservoir that contains only oil but no water!
1.4.6. Leverett J-Function

The capillary pressure function Pcap (Sw ) is often discussed in terms
of a dimensionless function known as the Leverett J-function. We
can see how this function arises by starting again with our bundle-
of-tubes model.
Recall that for the simplest form of the bundle-of-tubes model,
in which every tube has the same radius R, the capillary pressure is
given by Pcap = 2γow cos θ/R. In terms of the pore diameter, d, we
can say that Pcap = 4γow cos θ/d. But we also know that this bundle-
of-tubes model predicts that k = φd2 /96. So, the pore diameter can
be expressed as d = (96k/φ)1/2 . If we plug this into our equation for
Pcap , we find, after some rearrangement,

1 k 1
Pcap = √ . (1.4.22)
γ cos θ φ 6
The left-hand side of Eq. (1.4.22) is essentially a dimensionless

way of writing the capillary pressure function. For the bundle-
of-uniform-tubes model, the right-hand side is a constant, but we
know that this model is an oversimplification for real rocks. Moreover,
we already saw above that capillary pressure in a rock varies with the
saturation.
So, we can generalise Eq. (1.4.22) by replacing the constant on the
right by some dimensionless function of saturation, which Leverett
(1941) called the “J-function”. This function will be a property of
the rock, and its specific shape depends, in a complicated way, on the
details of the pore geometry. We then express the capillary pressure
in the form

1 k
Pcap = J(Sw ). (1.4.23)
γ cos θ φ
The logic that underpins the use of the J-function is that,

although properties such as k, φ and Pcap may vary throughout a
sedimentary unit, it is generally true that the J-function, as defined
by Eq. (1.4.23), is nearly invariant throughout the unit. Hence, if we
measure Pcap for one core, and convert it into a J-function, we can
then use Eq. (1.4.23) to estimate Pcap for other rocks in same unit.
Another use of the J-function is to take capillary pressure curves
that are measured in the lab and convert them into capillary pressure
curves for the reservoir. Assume that we measure Pcap in the lab using
two fluids that have certain values of γ and θ, say γlab and θlab ; then
Eq. (1.4.23) takes the form

1 k lab
P = J(Sw ). (1.4.24)
γlab cos θlab φ cap
In the reservoir, this rock would have the values of same k

and φ, but the fluids would be different, so there will be different
values of γ and θ, say γres and θres . Hence, in the reservoir we can
say that

1 k res
P = J(Sw ). (1.4.25)
γres cos θres φ cap
If we equate Eqs. (1.4.24) and (1.4.25), then we can rearrange and

say that

res lab γres cos θres
Pcap = Pcap . (1.4.26)
γlab cos θlab
Equation (1.4.26) shows how capillary pressures measured on a core

in the lab can be converted to the values that would occur in the
reservoir.
1.5. Two-Phase Flow and Relative Permeability

1.5.1. Concept of Relative Permeability
In Sec. 3, we defined and discussed the concept of permeability, in the
context of a rock that is fully saturated with a single fluid phase. But
we learned in Sec. 4 that reservoir rocks always contain at least two
fluid phases, oil and water, and sometimes three phases, oil, water
and gas. So, the concept of permeability must be extended to apply
to situations in which more than one phase is present in the pore
space.
The most obvious way to generalise Darcy’s law to account for
two-phase flow conditions is to assume that the flow of each phase
is governed by Darcy’s law, but with each parameter — pressure,
viscosity and permeability itself — being specific to the phase in
question, i.e.
−kw dPw −ko dPo
qw = , qo = , (1.5.1)
µw dx µo dx
where µw is the viscosity of water, kw is the effective permeability of
the rock to water, etc. The pressure in the oil phase and the water
phase differ from each other by the capillary pressure, which is a
function of the saturation.
However, it is more common to express the effective permeability
of the rock to water as the product of the single-phase permeability,
k (also known as the absolute permeability), and another parameter,
krw , known as the relative permeability of the rock to water ; likewise
for oil. Note that the relative permeability function is dimensionless.
Equation (1.5.1) can then be written as
−kkrw dPw −kkro dPo
qw = , qo = . (1.5.2)
µw dx µo dx
The relative permeabilities of each phase are functions of the phase
saturations. If part of the pore space is occupied by water, then
the ability of oil to flow through the pore space will obviously be
hindered, and vice versa. Hence, the relative permeability of a phase
will be a monotonically increasing function of the saturation of that
phase.
The precise shapes of these curves depend on the process that is

occurring. Specifically, they have a different shape during imbibition,
which is when a wetting phase displaces a non-wetting phase, than
they have during drainage, which is when the non-wetting phase
displaces the wetting phase.
We will first consider imbibition, such as occurs, for example,
when we inject water into a water-wet reservoir to displace the oil.
This process will start at Sw = Siw , where Siw is the irreducible water
saturation, which is the water saturation that remained in the rock
after oil had originally migrated into the reservoir. (The saturation
Sw = Siw is also precisely the point at which the capillary pressure
curve becomes unbounded.) Almost by definition, krw will be zero
when Sw = Siw . On the other hand, when Sw = Siw , kro will be
finite, but less than 1, as seen in Fig. 1.5.1.
Now imagine that we inject water into the rock, thereby increas-
ing Sw . Since the relative permeability of a phase is an increasing
function of the saturation of that phase, krw will increase, and kro
will decrease. This imbibition process will continue until we reach a
specific oil saturation, known as the residual oil saturation, So = Sor ,
at which kro has dropped to zero. This occurs at a finite value of
the oil saturation, not at zero oil saturation, as one might have
expected.
kro
krw
0
0 Siw 1-Sor 1
Sw
Figure 1.5.1. Relative permeability functions for water and oil.

These two values, Siw and Sor , are also known as the relative
permeability end-points, and the relative permeability values at these
saturations are known as the end-point relative permeabilities.
The precise shapes of the relative permeability curves depend on
the details of the pore structure of the rock. Power-law functions are
often found to be useful in fitting these curves. Note that relative
permeability curves are never linear functions of the saturation,
although this simple linear form is sometimes assumed, particularly
in fractured reservoirs, as data on the relative permeabilities of the
fractures is rarely available. The range of validity of the assumption
of linear relative permeability functions in fractured reservoirs has
been investigated by de la Porte et al. (2005).
1.5.2. Irreducible Saturations

The facts that Siw was not equal to zero after primary drainage
(when the oil first displaced the water to form the oil reservoir),
and Sor is not zero after imbibition (when water is injected to
flood the oil out of the reservoir) are of the utmost importance
in reservoir engineering, but it is not obvious that these residual
saturation values will not be zero. Also, in contrast to many
other rock properties, which can, at least partially, be understood
using the parallel-tube model of a porous rock, irreducible/residual
saturations cannot be explained by the parallel-tube model. In fact,
the parallel-tube model would erroneously predict that Siw = Sor = 0.
The phenomenon of irreducible/residual saturations is intimately
related to the heterogeneity and interconnectedness of the pores in
a rock.
We can gain a qualitative understanding of this phenomenon
using the simplest pore-space model that incorporates some degree
of heterogeneity and interconnectedness. Consider a pore doublet, as
shown in Fig. 1.5.2, in which one pore branches off into two pores of
different diameter, which then remerge to form a single pore:
Imagine that this doublet is initially filled with oil, as in
Fig. 1.5.2(a). We now slowly inject water from the left. According
to the Young–Laplace equation, Eq. (1.4.12), the capillary pressure
in each pore is inversely proportional to the pore radius, i.e. it is
oil
water
(a)
oil
water
(b)
oil
water
oil
(c)
Figure 1.5.2. A pore doublet used to illustrate how small isolated blobs of oil can
get trapped behind when water displaces oil in a reservoir, giving rise to a finite
value of the residual oil saturation.
proportional to R−1 . However, according to Poiseuille’s equation,

Eq. (1.3.13), the “permeability” of a pore is proportional to R2 . So,
from Darcy’s law, Eq. (1.3.3), the mean velocity in either pore is
proportional to R−1 × R2 = R. Hence, the water moves faster into
the larger pore than into the smaller pore, as shown in Fig. 1.5.2(b).
When the water in the larger pore reaches the end of the
doublet, it can enter the smaller pore from the far end, as in
Fig. 1.5.2(c), thereby trapping some of the oil behind it. This is
a simple demonstration of why the residual oil saturation, after the
imbibition of water, is not equal to zero.
1.6. Electrical Resistivity

The electrical resistivity of a rock is not a property that directly
affects oil production, nor does it appear in the governing equations
of fluid flow in a reservoir, but it is nevertheless very important

in reservoir engineering, because it can be measured in situ using
logging tools, and its value can then be used to infer the value
of the oil saturation. Practical issues related to the measurement
of resistivity using logging tools, and the interpretation of these
measurements, will be discussed in detail in the chapter on log
analysis. In the present chapter, we will discuss some of the basic
concepts and definitions related to the electrical resistivity of a fluid-
saturated rock.
The flow of electrical current is governed by Ohm’s law, which
states that the current, I, flowing through any conductor, is equal to
the voltage drop, ∆V , divided by the resistance, R:
∆V
I= . (1.6.1)
R
Electrical charge has units of coulombs, so current, which is the flow
of charge, has units of coulombs/second. Resistance therefore has
units of volt-seconds/coulomb, which are also known as ohms. In the
form of Ohm’s law given by Eq. (1.6.1), R will depend on the material
properties, but also on the shape and size of the conductor.
Now consider a cylindrically shaped conductor, of length L and
cross-sectional area A, as in Fig. 1.6.1. All other factors being equal,
the current will be proportional to A, and inversely proportional to L.
L
A
∆V
Figure 1.6.1. Cylindrical core with cross-sectional area A, length L, and axial
voltage drop ∆V . If the material has conductivity σ, then the current will be
given by I = σA∆V /L.
So, we expect that the resistance R can be expressed as

L
R=ρ , (1.6.2)
A
where ρ, the resistivity, is an intrinsic property of the material, and
does not depend on the geometry of the conductor. The resistivity
has units of ohm-meters. Hence, Eq. (1.6.1) can be written as
A ∆V
I= . (1.6.3)
ρ L
We can also define the conductivity as σ = 1/ρ, in which case we can
write Eq. (1.6.3) as
∆V
I = σA . (1.6.4)
L
In this form, it is clear that Ohm’s law is mathematically analogous to
Darcy’s law, with current (flow of electrical charge) being analogous
to fluid flow, voltage drop analogous to pressure drop, and electrical
conductivity analogous to the ratio of permeability/viscosity (i.e. the
mobility).
The electrical conductivities of the minerals that typically form
reservoir rocks are very low, as is the conductivity of hydrocarbon
fluids. But the water that partially fills the pore space of reservoir
rocks always contains salts such as NaCl or KCl, which render the
water conductive. The conductivities of these brines are typically 10
orders of magnitude higher than that of the rock minerals. Hence,
electrical current in a reservoir rock will flow mainly through that
portion of the pore space that is occupied by water.
As with permeability and capillary pressure, we can get some
idea of how the electrical conductivity depends on pore structure by
appealing to the bundle of parallel tubes model. Consider such an
idealised rock, as in Fig. 1.6.2, where A is the total area of the core,
and the An are the areas of the individual pores.
Imagine that the pores are all filled with a brine of conductivity
σw . If the core has length L into the page and the rock is subjected
to a voltage drop ∆V along its length, then the current through the
A1
A2
A3
Figure 1.6.2. Cylindrical tube model of a porous rock, used in developing a simple
model for the electrical formation factor.
nth tube will be

∆V
In = σw An . (1.6.5)
L
The total current is the sum of the currents through each tube:
N
N
N
∆V ∆V
I= In = σw An = σw An
n=1 n=1
L L n=1
∆V ∆V
= σw Apores = σw φA. (1.6.6)
L L
If we compare this with Ohm’s law, Eq. (1.6.4), we see that the
effective conductance of the fluid-saturated rock is σeff = σw φ. This
quantity depends on the rock and on the brine. However, we are not
really interested in the brine, so it would be preferable to extract
out a parameter that reflects only the properties of the rock. We
do this by defining — in general, independent of any pore geometry
model — the formation resistivity factor, also known as the formation
factor, as the ratio of the conductance of the brine to the effective
conductance of the brine-saturated rock:
σ(brine)
F ≡ . (1.6.7)
σ(brine-saturated rock)
The resistivity is the inverse of the conductivity, so we can also say
that
ρ(brine-saturated rock)
F ≡ . (1.6.8)
ρ(brine)
For our parallel-tube model, σeff = σw φ, and so F is predicted to be

equal to
σ(brine) 1
F ≡ = . (1.6.9)
σ(brine-saturated rock) φ
In contrast to the permeability, which has a strong dependence on
pore size, the formation factor, according to the bundle-of-tubes
model, has no dependence on pore size.
If we make the same argument as we did in the case of
permeability, i.e. only one third of the pores are aligned in the
direction of the voltage drop, then we would find
F = 3φ−1 . (1.6.10)
This parallel-tube model correctly tells us that F will be larger

for less porous rocks, but otherwise it is not accurate enough
for engineering purposes. Experimental measurements of F tend
to show a stronger dependence on porosity than the −1 power
that appears in Eq. (1.6.10). Archie (1942) proposed generalising
Eq. (1.6.10) by replacing both the factor of 3 and the exponent −1
with parameters that may vary from rock-to-rock. The result is the
famous “Archie’s law”:
F = bφ−m . (1.6.11)
The parameter b is often called the tortuosity, and m is called the

cementation index, but these names are outdated and not very useful.
Archie’s law in the form of Eq. (1.6.11) can fit many sets of
resistivity data consisting of different rocks from the same reservoir.
For sandstones, the exponent m usually lies between 1.5 and 2.5,
and is often close to 2; for carbonates, it can be as large as 4. The
parameter b is usually close to 1.0. Figure 1.6.3 shows some data on
Vosges and Fontainebleau sandstones, from Ruffet et al. (1991), fit
with b = 0.496 and m = 2.05.
Although Archie’s law is extremely useful in reservoir engi-
neering, it should nevertheless be remembered that it is not a
fundamental law of rock physics, but is merely a convenient curve-fit
that is usually sufficiently accurate for engineering purposes.
1000
100
F F = 0.496φ−2.05
10
1
1 10 100
Porosity, φ (%)
Figure 1.6.3. Measured values of formation factor on a set of Vosges and

Fontainebleau sandstones, as a function of porosity (Ruffet et al., 1991). Archie’s
law provides a reasonable fit, with b = 0.496 and m = 2.05.
It might appear that we could use Archie’s law to estimate

porosity, but there are much more accurate ways to estimate φ, as is
explained in the modules on core and log analysis. The usefulness
of Archie’s law, and of resistivity measurements in general, is in
estimating the water saturation. To understand how this is possible,
we need to consider rocks that are partially saturated with water,
and partially with oil.
For rocks that contain oil and water, we use the following
generalisation of Archie’s law, which is sometimes called Archie’s
second law :
F = bφ−m (Sw )−n , (1.6.12)
where n is called the saturation exponent. In water-wet rocks, n

is often close to 2, but in oil-wet rocks it may be as large as 10.
If one assumes that n is constant throughout a reservoir, or at
least throughout a certain rock unit, then Eq. (1.6.12) implies that
electrical resistivity measurements in a borehole can yield estimates
of the water saturation, and hence the oil saturation. This will be
discussed in detail in the module on well logging.
The scenario described above is more complicated in shaly sands.
In these rocks, the sand grains are coated by clay platelets. An ionic
electrical double layer then builds up along the surface of these plates.
This surface layer allows another path for current flow, separate
from the current flow through the brine-saturated pores that were
discussed above. To a good approximation, this surface current can
be thought of as being in parallel with the pore-current, and so
it adds an extra component to the conductivity of the rock. This
extra conductivity depends on the electrochemical properties of the
clays, but not on the intrinsic conductivity of the brine. The resulting
generalisation of Eq. (1.6.11) for shaly sands is
1
σ(brine-saturated rock) = (σw + BQv ), (1.6.13)
F
where Qv is the charge on the double layer, per unit volume, and B is
a constant. Equation (1.6.13), which is called the Waxman–Smits
equation, will be discussed further in the module on core analysis.
1.7. Fluid and Pore Compressibility

1.7.1. Fluid Compressibility
Quantities of oil are usually discussed in terms of barrels, which is a
measurement of volume. Likewise, gas is often measured in terms of
cubic feet. But the amount of volume taken up by a given mass of oil
will depend on the pressure to which the oil is subjected. When we
say that a well produces 100 barrels of oil a day, for example, we are
measuring these barrels at “atmospheric pressure”, which is 14.7 psi,
or 101 kPa in SI units.
The relationship between the volume and pressure of a fluid is
quantified in terms of the fluid compressibility, Cf , which is defined
as the fractional derivative of volume with respect to pressure (at
constant temperature, T , and for a fixed amount of mass):

1 ∂V
Cf = − . (1.7.1)
V ∂P T
Density is the inverse of specific volume, i.e. ρ = 1/v, where

v = V /mass, so it follows from the chain rule of calculus that the
compressibility can also be defined as

1 ∂ρ
Cf = . (1.7.2)
ρ ∂P T
The dimensions of Cf are 1/pressure, and so the units are 1/psi
or 1/Pa. The compressibility of a fluid usually varies with pressure,
but typical values of Cf are 0.5 × 10−9 /psi for water and about 1.0 ×
10−9 /psi for oil.3
1.7.2. Pore Compressibility

As oil flows from the reservoir to a well, two changes occur at any
given location in the reservoir: the pressure in the oil decreases, and
the pore space contains less oil. It is important to be able to relate the
change in the amount of oil stored in the pore space of the reservoir
to the change in the fluid (oil) pressure. This relation will obviously
involve the fluid compressibility, but also involves a property of the
rock known as the pore compressibility.
Consider a porous rock as shown below, with total (bulk) volume
Vb , pore volume Vp , and mineral grain volume Vm . Imagine that
the rock is compressed from the outside by a hydrostatic pressure
Pc , called the confining pressure. Inside the pore space is a fluid
at some pore pressure, Pp , which acts over the walls of the pores.
The confining pressure tends to compress both the bulk rock and
the pores, whereas the pore pressure tends to cause Vb and Vp to
increase. (The confining pressure Pc , which acts on the rock, should
not be confused with the capillary pressure Pcap , which acts within
the fluid.)
Recall from Eq. (1.7.1) that for a homogeneous solid or liquid of
volume V , subjected to a confining pressure P , the compressibility, C,
3
Note: when oil is taken from its pressurised state in the reservoir up to the
surface where it is at atmospheric pressure, any gas that had been dissolved in the
oil will be released, and the oil will actually shrink ! This phenomenon cannot be
described in terms of the compressibility of liquid oil; this shrinkage must be taken
into account when doing material balance calculations on a reservoir. However,
Eqs. (1.7.1) and (1.7.2) are applicable to the incremental pressure changes that
occur to the oil when it is flowing inside the reservoir.
Pc
Pc
Pp
Pc
Figure 1.7.1. Schematic diagram of a porous rock, subjected to an external

confining pressure Pc , and an internal pore pressure, Pp .
is defined as

−1 ∂V
C= . (1.7.3)
V ∂P
For a porous rock (Fig. 1.7.1), we need to consider two volumes,
the pore volume and the bulk volume, and two pressures, the pore
pressure and the confining pressure. So, we can define four different
compressibilities (Zimmerman, 1991):

−1 ∂Vb 1 ∂Vb
Cbc = , Cbp = , (1.7.4)
Vb ∂Pc Pp Vb ∂Pp Pc

−1 ∂Vp 1 ∂Vp
Cpc = , Cpp = . (1.7.5)
Vp ∂Pc Pp Vp ∂Pp Pc
The first subscript refers to the volume in question, either “bulk”

or “pore”, and the second subscript refers to the pressure that is
varying, either “confining” or “pore”. The pressures written outside
the derivative indicates that this pressure is held constant.
The pore compressibility with respect to changes in pore pres-
sure, Cpp , is useful in material balance calculations. The numerical
sum of the fluid compressibility and the pore compressibility, which
is known as the total compressibility, Ct = Cf + Cpp , appears in
the pressure diffusivity equation that is used in well-test analysis.
The bulk compressibility Cbc influences the velocity of seismic
compressional waves. The bulk compressibility Cbp is relevant to
subsidence calculations.
The numerical values of the various porous rock compressibilities

are controlled, as are all petrophysical properties, by the geometry
of the pore space. Roughly speaking, flat, crack-like pores are very
compressible, whereas fatter pores in the shape of circular tubes are
not very compressible.
The numerical values of the pore compressibility vary from one
rock type to another. Furthermore, the pore compressibility often
varies strongly with the pore pressure. Roughly, one can say that in
a sandstone reservoir, the pore compressibility Cpp is on the order of
about 1 − 10 × 10−6 /psi, or 2 − 15 × 10−4 /MPa. More details and
numerical values can be found in Compressibility of Sandstones by
Zimmerman (1991), and The Rock Physics Handbook by Mavko et al.
(2009).
References
Archie, G. E. (1942). The electrical resistivity log as an aid in determining
some reservoir characteristics, Petrol. Trans. AIME, 146, 54–62.
Bear, J. (1972). Dynamics of Fluids in Porous Media, American Elsevier,
New York.
Beran, M. (1968). Statistical Continuum Theories, Interscience, London.
de la Porte, J. J., Kossack, C. A. and Zimmerman, R. W. (2005). The
effect of fracture relative permeabilities and capillary pressures on the
numerical simulation of naturally fractured reservoirs. SPE Annual
Technical Conference held in Dallas, (SPE 95241).
de Marsily, G. (1986). Quantitative Hydrogeology, Academic Press,
San Diego.
Dullien, F. A. L. (1992). Porous Media: Fluid Transport and Pore Structure,
2nd ed., Academic Press, San Diego.
Leverett, M. C. (1941). Capillary behaviour in porous solids, Petrol. Trans.
AIME, 142, 159–172.
Mavko, G., Mukerji, T. and Dvorkin, J. (2009). The Rock Physics Handbook,
2nd ed., Cambridge University Press, Cambridge.
Ruffet, C., Gueguen, Y. and Darot, M. (1991). Complex conductivity
measurements and fractal nature of porosity, Geophysics, 56(6), 758–768.
Scheidegger, A. E. (1974). The Physics of Flow through Porous Media,
University of Toronto Press, Toronto.
Zimmerman, R. W. (1991). Compressibility of Sandstones, Elsevier,
Amsterdam.
Questions
1. Consider a reservoir that is shaped like a circular disk, 10 m thick,
and with a 5 km radius in the horizontal plane. The mean porosity
of the reservoir is 15%, the water saturation is 0.3, and the oil
saturation is 0.7.
(a) Ignoring the expansion of the oil that would occur when it is
produced from the reservoir, how many barrels of oil are in
this reservoir? One barrel = 0.1589 m3 .
(b) If the density of the oil is 900 kg/m3 , how much oil (in kg) is
contained in the reservoir?
2. In a laboratory experiment, a pressure drop of 100 kPa is imposed
along a core that has length of 10 cm, and a radius of 2 cm. The
permeability of the core is 200 mD, its porosity is 15%, and the
viscosity of water is 0.001 Pa·s.
(a) What will be the volumetric flowrate Q of the water, in m3 /s?
(b) What is the numerical value of q = Q/A, in m/s?
3. Imagine that the rock in problem 2 can be represented by the
parallel-tube model.
(a) Estimate the mean pore diameter, d, using Eq. (1.3.16).
(b) What is the mean velocity, v, of the water particles in the
rock?
(c) The importance of the inertia term, relative to the pressure
term in, say, Eq. (1.3.11), can be quantified by the Reynolds
number, defined as Re = ρvd/µ. What is the Reynolds number
in this experiment? Note that Darcy’s law is only accurate
when Re < 1 (Bear, 1972).
4. Consider a layered reservoir consisting of alternating layers, 1 m
thick, of rock 1, rock 2 and rock 3, where k1 = 1000 mD, k2 =
100 mD and k3 = 10 mD.
(a) What is the effective permeability of this rock, if fluid is
flowing parallel to the layering?
(b) What is the effective permeability of this rock, if fluid is
flowing perpendicular to the layering?
(c) Imagine that the reservoir consists of these three rock types,
in equal volumetric proportions, but occurring in a “random”
spatial distribution. Estimate the effective permeability in
this case.
5. Consider a small blob of oil surrounded by water. The surface
tension between the oil and water is 0.02 N·m. If the radius of the
blob is 0.05 mm, what is the value of the capillary pressure? Is the
pressure higher in the oil or the water?
6. Consider again the parallel tube model of a rock. Assume that the
diameter of each pore is 20 mm, γ = 0.02N·m, ρo = 900 kg/m3 ,
ρw = 1000 kg/m3 , the contact angle θ is 45◦ , and g = 9.8 m/s2 . If
this rock is placed in a tank containing water overlain by oil, as
in Fig. 1.4.7, to what height will the water rise in the pores?
7. Consider a homogeneous reservoir with φ = 0.20, k = 200 mD,
water–oil surface tension of γ = 0.03 N/m, and oil–water contact
angle of 35◦ . The oil density is 850 kg/m3 , and the water density
is 1050 kg/m3 . In the lab, we determine that the irreducible water
saturation occurs when the J-function is equal to 4.23. What will
be the height of the oil–water transition zone in the reservoir?
Hint: Use Eq. (1.4.23) to convert J(Swi ) to Pcap , and use
Eq. (1.4.21) on to convert Pcap to height.
Chapter 2
Introduction to Enhanced Recovery

Processes for Conventional Oil Production
Samuel C. Krevor∗ and Ann H. Muggeridge†

Department of Earth Sciences and Engineering,
Imperial College London,
Kensington, London SW7 2AZ, UK
∗
s.krevor@imperial.ac.uk
†
a.muggeridge@imperial.ac.uk
2.1. Introduction: Definition, Techniques and the

Global Role of EOR
2.1.1. The Aims of this Module
This series of lectures is designed to provide an introduction and
overview of enhanced oil recovery (EOR) technologies. It encom-
passes technologies applied today or at pilot scales in the past
towards the recovery of conventional, medium to light, crude oil.
The lectures will begin with a general overview of the role of EOR in
oil production today and a summary of the most recent forecasts
for the coming decades. Next, EOR will be described generally
within the framework of the goal of enhancing the recovery factor
by increasing microscopic and macroscopic displacement efficiencies.
The third section provides some detail about the working of gas
injection processes in the framework of fractional flow analysis to
evaluate improvement in displacement efficiency. The final section
includes a discussion of the more common chemical EOR techniques
including the emerging low salinity flooding technology.
47
2.1.2. Definitions and Techniques

EOR processes are oil recovery strategies that use the injection of
fluids, chemicals and heat into a reservoir that alter the thermo-
physical or chemical properties of the multi-phase fluid–rock system.
Table 2.1.1 shows broad categories of EOR technologies relative
to primary and secondary recovery techniques. There is still some
discussion about the precise definition (Hite et al., 2003), but most
find it useful to define EOR as those technologies that enhance
production through either the injection of heat or fluids into the
reservoir that were not initially present (Hite et al., 2003; Lake,
1989; Muggeridge et al., 2012). They are distinct from secondary
water injection and other pressure maintenance techniques. They
are often tertiary recovery processes, but there is no requirement
that this be the case and indeed most EOR techniques are most
effective if applied as secondary recovery mechanisms. The term
improved oil recovery (IOR), can encompass EOR techniques but
also includes reservoir management practices such as well placement
Table 2.1.1. EOR defined relative to other

production techniques.
Primary production
• Natural lift
• Artificial lift
Secondary production
• Waterflooding
• Pressure maintenance
EOR
• Thermal stimulation
• Chemical EOR
• Gas injection processes
Improved oil recovery
• Includes EOR
• New wells
• Reservoir management to increase sweep
Unconventional
• Shale oil — light tight oil
• Extra heavy crude
• Kerogen (oil shale)
Introduction to Enhanced Recovery Processes 49
and injection/production scheduling that increases production by

targeting bypassed regions of the reservoir. The term unconventional
production refers to production of hydrocarbons from tight reservoirs
where hydraulic fracturing is required.
A wide range of gas injection, chemical and thermal EOR
techniques have been used on commercial scales since the 1970s.
Their geographic deployment discussed in the next section reflects a
combination of local economic, geologic and regulatory circumstances
that favour a specific production technology. Table 2.1.2 shows
screening criteria for the deployment of EOR technologies in the three
broad categories. Gas injection can be applied across a wide range of
reservoir permeability but it is favoured in deeper reservoirs where
high pressures lead to miscibility with the resident hydrocarbon.
Gas injection targets medium to light crudes and can be economic
even when the remaining oil saturation is as low as 20%. Steam
injection by comparison is applied to heavy, unconventional oil. It
requires high oil saturations, reservoir permeabilities and deployment
in shallower reservoirs to be economic. On the other hand, it usually
targets heavy to extra heavy crudes that could not otherwise be
produced at all. Chemical EOR is a term used to describe techniques
that alter the water chemistry, either by adding artificial chemicals
(polymers, surfactants) or changing the salt content of the injected
water (low salinity or high sulphate). Polymers are typically used
for viscous oils in cooler, high permeability sandstone reservoirs as
long as the formation water is not too saline, although new polymers
are being developed that can target more challenging environments
(high salinity formation water, higher temperatures, carbonates).
Table 2.1.2. EOR Screening criteria (Taber et al.,

1997a, 1997b).
Method API Saturation Depth (m)
Nitrogen 35–48 40–75 >1,800

Hydrocarbon 24–41 30–80 >1,200
CO2 22–36 20–55 >800
Polymer >15 50–80 <2,700
Steam >8 40–66 <1,400
Low salinity water injection is only likely to be successful in sandstone

reservoirs where the formation water has a high salinity, the oil
contains polar compounds and there is a proportion of clay (kaolinite)
in the rock.
2.1.3. The Role of EOR in Current and Future

Global Oil Production
In 2014, EOR processes contributed 1.7 million barrels of oil per
day to global oil production (Koottungal, 2014; IEA, 2013b). This
represents about 2% of global oil production and was dominated
by production in four countries, the USA, Venezuela, Canada and
Indonesia (Fig. 2.1.1). In these countries, production from EOR
constitutes a larger fraction of the total relative to the world average.
EOR made up 10% of total production in the USA since 1992, 10%
of production in Canada and 16% in Venezuela in 2014.
The technologies applied for EOR are also distributed unevenly
around the globe. Steam injection contributes just over half of all
oil produced from EOR technologies whereas the injection of a
gas, predominantly CO2 or hydrocarbon fluids, contributes most of
the remainder. Chemical EOR processes have remained little used
Global oil production from EOR in 2014

Total production: 1.7 million barrels per day
Canada: 21%
Indonesia:
11% Brazil, Norway,
Venezuela: 22% Gas injection:
Turkey, UK, Thermal: 55%
Trinidad, 45%
Germany and
USA: 42% Egypt 4%
Figure 2.1.1. Oil produced using EOR techniques globally in 2014.

Source: Koottungal (2014).
× 10 5
Enhanced 8
production
(barrels per day)
6 Gas injection
4
Thermal
0
1992 1996 2000 2004 2008 2012
Figure 2.1.2. Oil produced using EOR techniques in the USA by technology since
1992.
Source: Koottungal (2014).
outside of China since the 1990s but have undergone a significant

amount of research and development. In the USA about 60% of the oil
from EOR came from gas injection and the remaining was produced
using thermal methods, nearly all steam injection. Additionally,
the fraction of oil recovered using gas injection has been steadily
increasing (Fig. 2.1.2). In Canada, by contrast, over 80% of enhanced
recovery is from steam injection and only 10% comes from gas
injection. In Venezuela, about two-thirds of the enhanced production
is from steam injection and the remaining third from hydrocarbon
gas injection. The production in Indonesia from EOR is entirely from
the largest steam flood project in the world, the Duri field. Steam
injection makes up most of the remaining 4% of EOR projects around
the world. The wide geographic deployment of steam injection is
indicative of those locations, particularly in Canada and Venezuela,
where there are significant resources of heavy crude oil. The relatively
limited geographic distribution of gas injection is in large part due
to the lack of large sources of CO2 outside of the Rocky Mountain
region of the Western USA.
Near term projections for oil produced through EOR show that it
will continue to play an important but limited role in oil production
through the middle of the 21st century. The most recent estimate
E. Europe/Eurasia
Africa 2012
2020
OECD Europe 2035
Middle East
Latin America
Asia Oceania
North America
0.2 0.4 0.6 0.8 1.0 1.2
(million barrels per day)
Figure 2.1.3. Modelled projections of oil production from all EOR operations
globally.
Source: (IEA, 2013b).
by (IEA, 2013b) shown in Fig. 2.1.3 projects that production from

EOR will increase only to 3.4 million barrels per day by 2035. In
the USA, projections estimate a doubling or tripling of oil produced
from CO2 injection, from around 250,000 barrels per day today
to somewhere between 450,000 and 1,000,000 barrels per day by
2040, depending primarily on the price of oil (USEIA, 2014; Wallace
and Kuuskraa, 2014). Based on information about specific projects
coming online before 2016 the study of (Wallace and Kuuskraa, 2014)
predicts that much of this increase will take place prior to 2020.
The limited application of EOR technologies projected is primarily
due to the abundance of oil still recoverable through secondary
production methods. However, it has also been impacted by the
rapid development of techniques for producing unconventional oil
and particularly shale oil, or so-called “light tight oil”. Production
from these reservoirs has a rapid rate of return, with payback on
the investment in as little as 2 years and thus favours development
by smaller companies with limited borrowing capacity (IEA, 2013a).
Thus, while in the USA total production from EOR is projected
to double slowly over 30 years, total oil production is expected to

have nearly doubled from 4.5 to 9.6 million barrels per day in the
decade 2010–2020. This is almost entirely due to production from the
Bakken, Eagle Ford and Permian basin tight oil formations (USEIA,
2014). Thus in the USA, EOR will play a diminishing role over the
next decade.
While EOR will be used to produce a small fraction of the
petroleum supply as long as low cost production technologies keep
up with demand, there is significant importance in the potential for
the deployment of EOR to increase ultimately recovered resources.
The average recovery factor for mature conventional crude projects
ranges from 20% to 40%. The promise of EOR is to increase these
recovery factors to the range 50–70% (Smalley et al., 2009). An
approximate understanding of the potential can be found as follows.
Various estimates place remaining oil in place at around 1013 barrels.
Currently an average recovery factor for fields is around RF = 0.25.
This suggests around 2.5 × 1012 barrels remain to be recovered
without EOR. An increase in the recovery factor by 10% points,
however, would result in 3.5 × 1012 barrels of oil to be recovered, an
increment of 1012 barrels of oil, which is nearly as much as has been
already produced historically! And a 50% increase in oil supply over
our previous estimate.
Various global estimates using analyses ranging in complexity
suggest that between 300 and 1,000 billion barrels of oil may be
incrementally recovered through the application of EOR worldwide
(ARI, 2009; IEA, 2013b; Wallace and Kuuskraa, 2014) (see Figs. 2.1.4
and 2.1.5). The US Energy Information Agency places oil supply
meeting at least 25 years of demand. The widespread deployment of
EOR extends this by at least 50% and thus it continues to be seen
as a technology waiting in the wings but of major importance.
2.2. Enhancing the Recovery Factor

2.2.1. The Recovery Factor
The recovery factor, RF , for a given oil field is the ratio of oil
produced, NP , to that estimated to be initially in place, N , both
Oil from CO2 High oil price

injection
Mbbl/day
Base scenario
Low oil price
Figure 2.1.4. Modelled projections of oil produced from CO2 -EOR in the USA.
Source: (USEIA, 2014).
Production cost Ultra

[USD/bbl] Arctic deepwater
Non-CO 2-EOR Oil shale

Extra heavy
oil
Other
Shale oil
conventional
Already MENA
produced conventional CO2-EOR
–
Technically recoverable resources (billion barrels)
Figure 2.1.5. Technically recoverable oil resources and associated production costs
in the absence of a CO2 price (non-CO2 -EOR) and in the presence of tax of
$150/tCO2 (CO2 -EOR).
Source: (IEA, 2013a).
measured at standard conditions

NP
. RF = (2.2.1)
N
Waterflooding is the most common recovery method for new fields.
Material balance and decline curve analysis are not generally useful
for estimating recovery factors fields where displacement is driving

production. Often the recovery factor is further broken into recovery
efficiencies attributable to particular aspects of the production
process. An approach used recently by BP and detailed in Smalley
et al. (2009) encompasses both technical and economic considera-
tions. The recovery factor is considered to consist of the product
of efficiencies in pore scale displacement, EP S , reservoir sweep,
Es , the connected volume factor, ED and an economic efficiency
factor EC ,
RF = EP S × ES × ED × EC . (2.2.2)
The application of EOR technologies is primarily to impact on either

or both pore scale displacement and the reservoir sweep. Improving
the connected volume factor is the domain of improved oil recovery
technologies and all of these combine to impact on the economic
cutoff.
2.2.2. Limits on Microscopic Displacement Efficiency

Subsurface permeable rocks have pore sizes of the order of 10−4 m
or less. When fluids are moving through constrictions at this scale,
interfacial forces between fluid phases and between the fluids and
the solid minerals of the pore walls are important and capillarity
governs the distribution of fluids at the pore scale. This is usually
characterised by the dimensionless capillary number,
vµ
Nc = , (2.2.3)
σ
where v is the the linear flow velocity, µ is the displacing fluid vis-
cosity and σ is the interfacial tension between the fluid phases. This
represents the balance between interfacial and viscous forces at the
pore scale and interfacial forces generally dominate in consolidated
rocks when Nc < 10−5 .
In this situation non-wetting fluids will move through larger pores

and through central parts of the pore space whereas wetting fluids
will move along the pore walls and inhabit smaller pores. In natural
rocks with mineral surfaces unaltered chemically by the presence of
hydrocarbons, aqueous fluids are the wetting phase relative to oil and
gas. In this case, oil moving out of the pore space of the rocks will
move through the central parts of the pore space whereas resident
brine can move freely along the walls of the pores. For this reason, oil
ganglia can be isolated when water films “snap off” pathways to flow
(Roof, 1970; Lenormand et al., 1983). These isolated oil ganglia are
the so called residually trapped non-wetting phase and the saturation
of residually trapped oil represents the limit on what can be displaced
when capillary forces dominate the microscopic fluid distribution. In
water-wet rocks, this is generally in the range of 15–35% of the pore
volume (Pentland et al., 2010) representing between a quarter and
one half of the initial oil in place (Fig. 2.2.1). The oil saturation in
these systems is generally reduced to the residual saturation once a
AI-Mansoori et al. (2009) Crowell et al. (1966) Geffen et al. (1952) Jerauld (1997)
Kleppe et al. (1997) Kralik et al. (2000) Land (Berea) (1971) Land (Alundum) (1971)
Ma and Youngren (1994) McKay (1974) Pickell et al. (1966) Suzanne et al. (2003)
70
60
50
S(nw)r (%)
40
30
20
10
0
0 10 20 30 40 50 60 70 80 90 100
S(nw)i (%)
Figure 2.2.1. Compilation of initial residual trapping curves in the literature.

Source: Pentland et al. (2010).
single pore volume of water has swept through the system (Salathiel,
1973; Anderson, 1987).
The presence of hydrocarbon alters the wettability of the local
mineral surface to be oil wetting. Because the initially non-wetting
phase will have originally inhabited the larger pores, this often leads
to a scenario where the rocks are oil wet in the larger pores and water-
wet in the smaller pores. This is referred to as a mixed wet system and
it is thought to be the prevailing scenario in oil reservoirs. In mixed
wet systems nearly all of the oil can be displaced but this requires
many more pore volumes of a displacing, generally wetting, fluid to
be injected than production from water-wet systems (Salathiel, 1973;
Anderson, 1987).
The residual saturation is stable for a given rock so long as
capillarity dominates the pore scale force balance of the fluids in the
pores, i.e. Nc < 10−5 . A number of EOR techniques are understood
to work primarily through the disruption of the role of capillarity
in determining the local fluid distribution (Fig. 2.2.2). So-called
chemical techniques use aqueous fluids for the displacement of oil
but with dissolved solutes that dramatically lower the interfacial
Wagner & Leach

1.0
Do
Fo
M
mbr
oo
ste
Sor /Sor, wf
re
r
Ab loba
owsk
Du
&
ra
Pre
S
y
s
i&
0.5
Tab
d
Bro
er
Du P
wn
el l
rey
Nonwetting Residual
Wetting Residual
0
10–8 10–6 10–4 10–2 100
Capillary Number, Nc
Figure 2.2.2. Residual saturation decreasing after threshold capillary numbers

are exceeded.
Source: Fulcher et al. (1985).
tension between the water and the oil. Solvent flooding uses a
displacement fluid that is miscible with the hydrocarbon phase, such
as CO2 or other hydrocarbons, thereby picking up the residual oil
and recovering it with the injected solvent.
An approximate analysis can provide an estimate of how much
the interfacial tension needs to be reduced to result in an effect.
Typical interstitial field velocities are v ≈ 1 m d−1 ≈ 10−5 m s−1 .
The interfacial tension between oil and water is generally σ =
6 × 10−2 N m−1 and the viscosity of light crude oil is similar to
water, µ = 10−3 Pa·s. This results in Nc ≈ 10−7 which suggests
that increases of a factor of 100 or more are required to initiate
desaturation of the residual non-wetting phase. Thus EOR techniques
focused on reducing interfacial tension require chemicals that can
bring interfacial tension to less than 10−3 N m−1 . It is also worth
noting that a displacement at these capillary numbers also has
a large impact on the other multi-phase flow properties. Relative
permeability curves begin to increase and capillarity significantly
weakens (Fig. 2.2.3).
The overall effect of enhancing recovery through an increase in
the microscopic displacement factor is to shift the range of values
0.5 < Eps < 0.75 to 0.7 < Eps < 1.
1.0 100
IFT(mN/m)
mo/mw = 1.5 s = 0.1
mo/mw = 1.38 s = 0.3
Water Relative Permeability (Krw)
0.8 80
Oil Relative Permeability (Kro)
mo/mw = 1.28 s = 34
Oil Recovery, Percent. PV
0.6 60
0.4 40
Legend
0.2 20 s = 34 mN/m
s = 0.03 mN/m
s = 0.01 mN/m
0.0 0
0.0 0.2 0.4 0.6 0.8 1.0 0 1 2 3 4 5
Water Saturation, SW Pore Volumes Injected, Vpi
Figure 2.2.3. The impact of interfacial tension reduction on relative permeability

and oil recovery in a linear core flood.
Source: Amaefule and Handy (1982).
2.2.3. Limits on Macroscopic Displacement Efficiency

The oil recovery factor obtained from a reservoir is always less than
the microscopic displacement efficiency because the injected fluids
only pass through some of the reservoir volume, even when there
is good communication between the injection and production wells.
This reduction in recovery is quantified by the macroscopic sweep
efficiency and is a consequence of the properties of the oil and the
injected fluids as well as the geology of the reservoir formation. The
underlying causes are classified into three types:
(1) Heterogeneity in the fluid movement due to viscous instability;

(2) Vertical segregation of the fluids due to buoyancy effects;
(3) Heterogeneity in the reservoir rock permeability leading to
channelling.
Viscous instability refers to the development of viscous fingers as

shown in Fig. 2.2.4. These occur when the displacing fluid is less
viscous than the oil, e.g. when gas displaces a light oil or water
displaces a viscous oil. They are initiated by small scale and low
level permeability fluctuations in the rock. These perturb the front
between the oil and displacing fluid. The perturbations then grow
because of the lower viscosity of the displacing fluid. It is important
L
x=
2 w
xf fluid Flo
x= d
p lace
Dis
1
uid
0 ing fl
x= lac
p
Dis ε
α = dip angle
Figure 2.2.4. Conditions for viscous instabilities to form.

Source: Lake (1989).
to remember that the exact pattern of viscous fingers that will form
in a given displacement is impossible to predict (they are random)
although it is possible to predict their average behaviour (e.g. Todd
and Longstaff, 1972; Koval, 1963; Homsy, 1987). This contrasts with
channelling caused by rock heterogeneity in which the displacing
fluid always follows the same path through the rock, the path being
determined by the permeability pattern.
Viscous fingers are most likely to form in miscible displacements
although they can also form in immiscible displacements, provided
that the viscosity ratio between the displacing fluid and the oil
is sufficiently high. A formal derivation of this condition can be
obtained by considering small perturbations to the location of a
displacement front, Fig. 2.2.4. In particular, one derives expressions
for the movement of the bulk front, dxf /dt, and compares that with
the movement of the perturbation, d(xf + )/dt. The condition for
instability is given by,
d d(xf + ) dxf
= − > 0. (2.2.4)
dt dt dt
We do not have time to cover the derivation in this module,
but a remarkably simple condition falls out of this analysis. In
miscible displacements, ignoring gravity, the condition for stability
is dependent on the viscosity ratio
µo
M= ≤ 1. (2.2.5)
µd
To determine whether viscous fingers are likely to form (where µo is
the oil viscosity and µd is the viscosity of the miscible displacing
fluid). In immiscible displacements the condition for stability is
dependent on the shock front mobility ratio,
kr,w (Swf ) k (S )
µw + r,oµo wf
Msf = kr,o (Swc )
≤ 1, (2.2.6)
µo
where krw (Swf ) and kro (Swf ) are the water and oil relative perme-
abilities at the shock front saturation, Swf , and kro (Swc ) is the oil
relative permeability at the connate water saturation.
Viscous fingers will form if M or Msf > 1 and the higher the
mobility ratio, the lower the sweep (Fig. 2.2.5). It is important to
remember that a 1D Buckley–Leverett analysis does not capture
the effect of macroscopic viscous fingering even if there is a large
viscosity ratio between the oil and the displacing fluid as viscous
fingering is a 3D phenomenon. It should also be noted that it is
very hard to explicitly model viscous fingers using conventional
reservoir simulators. You need a very fine grid and random perme-
ability or saturation fluctuation to trigger the fingers (Fig. 2.2.6).
It is more usual to capture the average effects of viscous fingers
using an empirical model such as the (Todd and Longstaff, 1972)
model.
100 M = 0.216 M=1

M
2.95
90 4.62
8.55
17.3
Breakthrough
80 38.2
Area Contacted by Drive, % Total Area
71.5
70
60
50
40
30
20
10
0 1.0 2.0 3.0

Pore Volume Injected
Figure 2.2.5. The impact of viscous instability on sweep for a miscible displace-
ment in a quarter five spot pattern.
Source: Haberman (1960).
M= M= M=
PVI = . PVI = . PVI = .
Oil saturation
Figure 2.2.6. Viscous fingering reduces the aerial sweep in a Hele Shaw cell.
Source: Djabbarov (2014).
Gravity can enhance or stabilise viscous fingering depending

upon whether it acts to increase the mobility of the displacing fluid
(e.g. injecting gas underneath oil to displace the oil upwards) or to
decrease its mobility (e.g. injecting gas on top of oil to displace the oil
downwards). In the latter case, viscous fingers may still form unless
the injection rate is sufficiently slow that gravity dominates the flow.
For miscible gas injection this condition occurs when (Dumore, 1964)
ρo − ρg
u< kg, (2.2.7)
µo (ln M )
where u is the injection velocity, ρo is the oil density, ρg is the
gas density, k is the permeability and g is the acceleration due to
gravity. An approximate analysis also shows that velocities here
are too small to be useful. Take ∆ρ = 400 kg m−3 , k = 10−14 m2 ,
µo = 10−3 Pa·s and M = 10 and this results in u < 1.84×10−3 m d−1 .
Decreasing the mobility ratio to M = 1.1 only improves this to
u < 4.4 × 10−2 m d−1 .
Thus in most cases this stable velocity results in an injection rate
(and hence a production rate) that is so low that it is uneconomic,
even if the reservoir structure allows the gas to be injected above
the oil in this way. Instead engineers tend to focus on reducing
the viscosity ratio between the injected fluid and the oil (polymer
injection) or reducing the mobility of the injected fluid by, for
example, injecting water and gas either simultaneously or in slugs

(water alternating gas, WAG).
Gravity can also reduce macroscopic sweep if the injected fluid
has a large density difference with the oil as well as a lower
viscosity and is injected horizontally (Fig. 2.2.7). In this case there
will be gravity segregation of the fluids (e.g. Dietz, 1953; Christie
z Gas s
q
α
h h
+ ∆h
∆ Oil
ht
l
Θ
i) Stable case (M = 0.5)

th
Dep
Leng
th
ii) Conditionally stable case (M = 10)

th
Dep
Leng
th
iii) Unstable case (M = 10)

th
Dep
Leng
th
Sheldon and Fayers

High resolution simulation
Figure 2.2.7. Conditions for gravity override in a homogeneous, but anisotropic

system.
Source: From Fayers and Muggeridge (1990).
et al., 1990; Fayers and Muggeridge, 1990; Tchelepi and Orr, 1994).
This is particularly important in gas injection processes and CO2
sequestration when the injected gas migrates quickly to the top of
the reservoir and then flows rapidly along the top, forming a gravity
tongue (Fig. 2.2.7). This effect may be less in a dipping reservoir
if the gas is injected at the top of the dipping layer and if the
vertical permeability is low. It is possible to estimate whether a
gravity tongue will form using the dimensionless number (Fayers and
Muggeridge, 1990).

u(1 − 1/M )µo h
Rv/g = 2 −θ , (2.2.8)
∆ρgkz L
where ∆ρ is the density difference between the oil and the gas (kg
m−3 ), kz is the vertical permeability (m2 ), h is the reservoir thickness
(m), L is the well spacing (m) and θ is the dip (radians). Note, this
can only reliably calculate the effects of dip when θ < 0.1 (about 10◦ ).
The displacement will always be stable if Rv/g < 0 (because M < 1
or because of a large dip). A gravity tongue will form if 0 < Rv/g <
1, for 1 < Rv/g < 10 flow will be influenced by both gravity and
viscous effects whilst the displacement will be viscous dominated if
Rv/g > 10. It should be noted that it is very difficult to use numerical
simulations to model unstable gravity tongues because a high degree
of vertical refinement is needed to capture the leading edge of the
tongue, instead it may be better to use a vertical equilibrium option
(if available in your simulator) if the above calculation suggests flow
will be gravity dominated. As with viscous instability the main way to
manage gravity segregation is through control of the mobility of the
injected fluid, typically this is via water alternating gas injection. The
water tends to reduce the mobility of the gas and, even if it does not
fully prevent the formation of a gravity tongue, it will improve sweep
as the water will tend to sink towards the bottom of the reservoir
whilst the gas rises to the top.
Geological heterogeneity is perhaps the most usual cause of
reduced macroscopic sweep although it can, on occasion, improve
sweep e.g. a fining upwards sequence will reduce the tendency of
gas to form a gravity tongue above oil during gas injection. EOR
processes are thought to be more sensitive to geological heterogeneity

because of the larger viscosity ratios between the injected fluid and
the oil. This results in a higher degree to channelling through larger
scale heterogeneities than would be seen in an ordinary waterflood.
Smaller scale heterogeneity is believed to increase dispersion on the
field scale (Mahadevan et al., 2003) which can further reduce the
effectiveness of EOR processes by diluting the active agent in
the injected water (e.g. polymer or low salinity water) or preventing
miscibility developing between injected gas and the oil. Conversely
transverse dispersion can reduce the level of viscous fingering and
thus improve sweep.
A major issue when attempting to quantify the impact of
geological heterogeneity on macroscopic sweep is uncertainty. In most
reservoirs even though geologists may have a good understanding
of the depositional environment in which the reservoir rocks were
laid down and there is data regarding the permeability distribution
along the wellbore we simply do not know the exact permeability
distribution between wells. Technically it should be possible to
create a large number of possible realisations of this permeability
distribution, conditioned to well data, but in practice there is often
neither time or resources to be able to simulate flow through all these
realisations and thus evaluate the range of possible outcomes from a
chosen EOR process.
Unlike for viscous instabilities or gravity influenced flows, there
are also no reliable dimensionless numbers that can be used to
estimate the impact of heterogeneity on performance. Some authors
advocate the use of the (Dykstra and Parsons, 1950) coefficient.
This assumes that the reservoir is layered but has been extended
for non-layered systems by (Jensen and Currie, 1990).
Vdp = 1 − e−Sk ,
whereas,

1
n
1
Sk = 1+ (ln ki − ln k),
4(n − 1) n−1
i=1
for a log normal permeability distribution. The Lorenz coefficient

(Schmalz and Rahme, 1950) is another widely used measure of
heterogeneity but, to date, there is no robust heterogeneity index
that both correlates with EOR performance in all types of hetero-
geneous reservoirs and can be calculated without resorting to a flow
simulation.
The usual way to improve macroscopic sweep, if oil has been
bypassed on the interwell scale and can be identified by seismic
surveying, is via the drilling of infill, horizontal or multilateral
wells. For smaller scale heterogeneities, flood front conformance
can be temporarily improved by shutting off high permeability
zones using mechanical means or polymer gels e.g. (Seright et al.,
2003), thereby forcing the injected fluid into lower permeability
zones and thus displacing additional oil. This type of treatment is
most effective if the high permeability zone is physically isolated
from these lower permeability zones by continuous, impermeable
zones (e.g. shales). If this is not the case then the injected fluid
will simply flow around the gel and back into the higher perme-
ability zone further into the reservoir (Sorbie and Seright, 1992).
Mitigating the adverse effects of geological heterogeneity on EOR
and, in particular, reducing the uncertainty in what these effects
will be remains one of the most significant challenges for reservoir
engineers.
In summary, macroscopic sweep can be improved through mobil-
ity control, either of injected water into particularly viscous oil or in
the control of gas being injected for miscible displacement floods. In
the latter case, both viscous instability and gravity override can be
controlled through the application of WAG. The promise of EOR in
this area is to improve values of macroscopic sweep from ES = 0.6
to upwards of ES = 0.8. Combining enhancements in pore scale
displacement and macroscopic sweep efficiencies improves recovery
factors from at least RF = 0.3–0.8. This is promising on the one hand.
On the other, this does not consider the impacts of economic cutoffs
and compartmentalisation. Thus recoveries achieving RF > 0.6, such
as the Prudhoe Bay field off the N . Slope of Alaska are impressive

indeed.
2.3. Gas Injection

While the injection of water to maintain pressure in a reservoir
and displace oil is an effective recovery mechanism, as discussed
in Sec. 2.2, capillary trapping results in maximum microscopic
displacement efficiencies of only 50–75% of the oil initially in place.
The injection of fluids that are miscible with the oil phase, however,
means that the microscopic displacement efficiency can be reduced
to a complete recovery of the oil. This is the basis for solvent gas
injection EOR, where the term “gas” here is used loosely to refer
to a light non-polar fluid phase such as CH4 or CO2 that are gases
at atmospheric conditions. At the depths at which gas injection is
generally applied, these fluids, and CO2 in particular, are not in the
gaseous state.
Quantitative descriptions of chemical component transport, fluid
phase displacement and fluid phase chemical composition are the
aims of a study of gas injection processes. The displacement processes
can involve up to three fluid phases and require representation of
three or four constituent components that can transfer between the
phases. This makes their description highly complex and generally
within the domain of numerical simulation. Detailed references on
gas injection processes include (Lake, 1989) and (Orr, 2007) and
the latter provides a comprehensive set of analytical solutions to
two-phase multi-component displacements. In this lecture, we will
only learn some of the basic concepts underlying the physics of gas
injection. I will first discuss qualitatively the representation of mis-
cibility on ternary phase diagrams. A brief refresher of the Buckley–
Leverett formulation of the flow equations for two-phase immiscible
displacement will be used as a starting point to develop the equations
for a two-phase multi-component system with miscibility. These tools
will then be applied in one of the simpler examples of a miscible
displacement process, a 1D first contact miscible water alternating

gas injection problem.
2.3.1. Phase Equilibrium for Gas Drives

When a gas like CO2 or CH4 is injected into an oil phase a
number of chemical component transfers will take place between
the fluids. Parts of the oil will vapourise into the gas-rich phase
and parts of the gas phase will dissolve into the oil. This changes
the chemical composition of each fluid phase and as each phase
contacts “fresh” oil or gas a new set of chemical transfers will take
place changing the composition even further. These stages of phase
contact, chemical equilibrium, movement, contact and reequilibrium
have major impacts on the flow of each through the controlling
relationship between the chemical composition of a phase and its
saturation.
The relationship between phase saturation and chemical compo-
sition for oil–gas systems is conventionally illustrated on a ternary
diagram. We will approach this in two parts — first, we will discuss
the use of a ternary diagram to represent the overall composition
of a system in terms of three chemical components. Then, we will
separately introduce the relationship between chemical component
composition and phase equilibrium in simpler, two component
systems. Finally, we will bring the two concepts together in a three-
component two-phase ternary diagram.
Figure 2.3.1 shows a ternary diagram for a single-phase system.
The vertices of the diagram are the three chemical components.
Conventionally a lighter component, C1 is the top vertice, an
intermediary component, C2 is on the right and a heavy component
C3 is on the left. For an example with a real system, see Fig. 2.3.8
and references (Orr et al., 1981; Gardner et al., 1981). Often the
intermediate and heavy components are pseudo components, an
arbitrary division of the crude oil into light and heavy fractions.
A fluid phase made up of various mole fractions of each component
is represented by a point within the triangle. The lines in the triangle
represent lines of constant mole fraction for one of the components.
C
.
.
. .
.
. .
. . .
. .
C C
Figure 2.3.1. A single-phase ternary diagram. A fluid with composition a is

composed of 60% C1 , 30% C2 , and 10% C3 .
Thus, a fluid with composition a is made of 60% C1 , 10% C3 and

30% C2 . Can you spend a few minutes right now and mark in the
notes the following compositions:
• b — 20% C1 , 50% C2 and 30% C3 ;

• c — 50% C1 , 20% C2 and 30% C3 ;
• d — 30% C1 , 10% C2 and 60% C3 .
When multiple phases exist there will be a region of the diagram in

which the overall chemical composition can only be represented by
the sum of the chemical compositions of two coexisting but separate
phases.
Consider first, however, a pressure-composition diagram for a
two-component system, CO2 and n-hexane at 333.15 K, shown in
Fig. 2.3.2. In this system, there are potentially two phases, a liquid
hexane rich phase and a gaseous CO2 -rich phase, and there are two
components, hexane and CO2 . If we refer to CO2 as component one,
C1 , and hexane as component two, C2 , we can see that C1 = 1 − C2 .
Similarly, if we refer to the liquid phase as phase one and the
gaseous phase as phase two, the respective saturations, Si are related
Figure 2.3.2. Vapour–liquid equilibria for mixtures of CO2 with n-hexane at

333.15.
Source: Potoff and Siepmann (2001).
through S1 = 1 − S2 . You can see on the diagram that to the

left of the bubble point curve, S1 = 1 and to the right of the
dew point curve S2 = 1. Within the envelope, the two phases are
in equilibrium although the specific mole fractions of each phase
are dependent on the overall chemical composition of the system.
Consider a fluid mixture with overall chemical composition within
the two-phase envelope. The chemical composition of the liquid part
of the mixture is given by the bubble point chemical composition,
c11 , the concentration of chemical component one in phase one. The
chemical composition of the vapour part of the mixture is given by
the dew point chemical composition, c12 , the concentration of the
chemical component one in phase two. The constraints of the specific
liquid and vapour chemical compositions along with the total overall
concentration of the chemical components results in a determination
of the saturation of liquid and vapour. In other words, the total
concentration of component one is given by,
C1 = c11 S1 + c12 S2 = c11 S1 + c12 (1 − S1 ). (2.3.1)
Light component, injection

C fluid, e.g. CO2, CH4
Phase 1: C rich
c , c ,c
a Plait point
c ,c ,c
Single-phase
Phase 2: C rich
C C
Heavy pseudo component, Crude oil composition Intermediate pseudo
e.g. C7+ component, e.g. C6–
Figure 2.3.3. A two-phase ternary diagram. A fluid mixture with overall com-
position a will exist as a mixture of two phases with component concentrations
given by the ends of the tie line on the binodal curve.
Thus the saturation is determined by,

C1 − c12
S1 = . (2.3.2)
c11 − c12
A similar analysis applies to a three component system with two
phases represented in a ternary diagram. In the ternary diagram
in Fig. 2.3.3, a phase rich in the light component is fully miscible
with a phase rich in the heavy component and the same is true
between the intermediate and heavy components. There is a region
of immiscibility, however, between phases rich in the light and
intermediate components. An overall fluid mixture composition in
this two-phase region given by the point a would be made of a
corresponding amount of C1 , C2 and C3 . No single-phase with
this particular composition is stable on this diagram and thus C1
would represent the sums of the concentration of component 1 in
phase 1, c11 , multiplied by the saturation of phase one, S1 , and the
concentration of the component in phase two, c12 , multiplied by the
saturation of phase two, S2 ,
C1 = c11 S1 + c12 S2 = c11 S1 + c12 (1 − S1 ). (2.3.3)
As with the binary component system, there is a line going through

the overall chemical component composition and intercepting the
phase boundaries at the locations where the chemical composition of
the phases is in equilibrium with the the total chemical composition.
In this case, however, it is not just a function of pressure, but also
the total chemical composition. In any case, for any given ternary
chemical composition, C1 , C2 and C3 within the two-phase envelope,
there are specific phase chemical compositions, c11 , c21 , c31 , c12 , etc.
associated with that overall composition. The line connecting the
phase boundaries with the chemical composition within the two-
phase envelope is known as a tie line.
Similar equations could be written for components two and three.
Note that all ternary diagrams only apply at a single pressure and
temperature, and the compositions of the fluids at the end of the
tie lines, e.g. c11 , c12 for component C1 , are a function of the total
mixture concentration only, C1 , C2 , C3 at this fixed pressure and
temperature. This is determined by the rules of thermodynamic
equilibrium of the two phases and thus the chemical composition
of each phase, e.g. c11 and c12 , can be thought of as constants
determined by the total chemical composition at the given pressure
and temperature.
An important result of this is that the respective saturations of
the two phases is determined by the total chemical composition. This
is seen by rearranging Eq. (2.3.3),
C1 − c12
S1 = . (2.3.4)
c11 − c12
The saturation could be equivalently found through similar equations
in terms of C2 or C3 . In other words, on a given tie line, the saturation
of the fluids is entirely determined by one total chemical component
concentration.
Only a few potential tie lines are shown in Fig. 2.3.3, but
there will be a tie line associated with any composition inside
the two-phase envelope with corresponding phase compositions and

saturations.
The other significant point highlighted on the figure is the Plait
point. This is the point at which the length of the tie lines vanish
to zero. One way to think about this is in terms of “developing
miscibility”. A phase initially rich in C1 such that it coexisted with a
phase rich in C2 could be diluted with contributions of the component
C3 until the total composition moved out of the two-phase region and
the fluids were miscible. This turns out to be a key process during
gas injection that we will now explore further.
Consider first a gas injection process represented by the ternary
diagram shown in Fig. 2.3.4. The resident oil has a composition given
by the point on the lower right of the diagram and the injection
gas is entirely composed of component C1 . This gas is fully miscible
with the resident oil because the mixing line between the two end
member chemical compositions (injected gas and resident oil) does
not pass through the two-phase region. This process is called first
contact miscible because the injected gas is already miscible with the
resident oil upon first contact between the two fluids.
Mixing
line
Resident
oil
C C
Figure 2.3.4. A two-phase ternary diagram showing the mixing line for the
injection of C1 , first contact miscible with the resident oil.
100 Vol % CO2
Single Contact Points

UL Multiple Contact Points
Plait Point
Plait Point
l
ica
rit
LL fC
O e
n
io Lin
e ns Tie
Ext
100 100
Crude
Vol % Vol %
C+2 C–6
100 Vol % CO2
Single Contact Points

Plait
Point
Li f
Ti n O
ne
al o
ic si
e
rit ten
Ex
100 100
C
Crude
Vol % Vol %
C+2 C–6
Figure 2.3.5. Ternary diagrams of mixtures of CO2 with Wasson crude at 1,350
psi (above) and 2,000 psi (below). Both are at 105◦ F. Increasing pressure results
in a larger field of potential miscibility with the injected fluid.
Figure 2.3.5 shows a ternary diagram with two-phase regions

for a mixture of CO2 with Wasson crude oil at two pressures
and 105◦ F.
In the case that the injected fluid is not initially fully miscible
with the resident oil, there are dilution pathways along which the
fluids can develop miscibility. These are called multi-contact miscible
processes for reasons that will become apparent shortly. Consider
Va mpos
co
po itio
uri
sin n
B
gg
as
A D
Plait
C
E point
Miscible
Resident
oil
C C
Figure 2.3.6. A two-phase ternary diagram showing the development of miscibility
through a vapourising gas drive, injecting C1 , multi-contact miscible with the
resident oil.
first a system where miscibility is developed as the injected gas

incorporates increasing amounts of heavier elements of the resident
oil into the gas-rich phase. This is shown in Fig. 2.3.6. Injecting a
gas made entirely of component C1 into the resident oil results in
a mixture at point A that falls within the two-phase region. The
mixture quickly equilibrates to form a gas-rich phase with a new
composition given by point B and an oil-rich phase on the opposing
end of the tie line. The new gas-rich phase moves further to contact
more fresh resident oil and forms a mixture with overall chemical
composition given by point C on the diagram. This again falls in
the two-phase region resulting in the formation of a gas-rich phase
now with the composition given by point D. You can see from the
diagram that as the gas-rich phase continues to contact more of
the uncontacted resident oil, it continues to pick up amounts of C2
and C3 such that the gas phase composition evolves until miscibility
has been achieved. Because of the multiple contacts between newly
equilibrated gas phase and unconnected oil, this is known as multi-
contact miscibility. In addition to this being a multi-contact miscible
Injected
gas
A C E
B
D
Miscible
Resident Condensing oil
oil composition
C C
Figure 2.3.7. A two-phase ternary diagram showing the development of miscibility

through a condensing gas drive, injecting a fluid with components C1 and C2 ,
multi-contact miscible with the resident oil.
process, this particular example is known as a vapourising gas drive

because the mechanism for the development of miscibility is to
increasingly vapourise components of the oil into the gas-rich phase.
There is another mode of multi-contact miscibility known as
the condensing gas drive. Consider the gas injection process shown
in Fig. 2.3.7 where a gas, now containing fractions of components
C1 and C2 , is injected into the resident oil with a composition
on the lower left side of the ternary diagram. The first contact
produces a mixture in the two-phase region given by the overall
composition at point A. This equilibrates to two phases with chemical
compositions given by the ends of the tie lines. In this scenario, the
gas-rich phase is moving away from miscibility but the oil-rich phase,
condensing components from the gas, is moving towards miscibility
as it vapourises the oil components. In other words, the equilibrated
oil with chemical composition at point B is closer to miscibility with
the injected gas. As more gas is injected, the oil with this composition
will again mix forming a mixture with composition at point C. This
mixture will separate into two phases and the oil rich component will
have a chemical composition given by point D, closer to miscibility

with the injected gas. Eventually, the oil-rich phase will condense
enough components out of the gas such that it has miscibility with
subsequently injected gas.
To generalise this discussion, consider the potential gas injection
scenarios possible with gas compositions g1 or g2 and reservoir oil
compositions r1 or r2 shown in Fig. 2.3.10. Take a few minutes to
answer the following questions using the diagram.
Classify the following injection scenarios as first contact miscible,
a vapourising gas drive, a condensing gas drive, or immiscible. If this
is a a multi-contact miscible process, show some of the steps to the
development of miscibility.
• The injection of gas with composition g2 into the oil with
composition r2 ;
• The injection of a gas with composition g1 into the oil with
composition r2 ;
composition r1 ;
composition r1 .
The injection of gas with composition g2 into the oil with
composition r2 would be a first contact miscible process because the
mixing line between the fluids does not pass through the two-phase
region. The displacement of reservoir fluid r2 with gas of composition
g1 would be multi-contact miscible and result in a vapourising gas
drive as the gas-rich phase approaches miscibility with the resident
oil. Similarly the injection of gas with composition g2 into oil with
composition r1 would be multi-contact miscible and result in a
condensing gas drive as the oil-rich phase approaches miscibility
with the injected gas. If you considered the injection of gas g1 into
reservoir oil with composition r1 and followed the same procedure
as for the previous multi-contact miscible systems you would find
that miscibility would not be approached for this system. This would
thus be classified as an immiscible flood. In order for miscibility to
develop, the initial compositions of the gas and oil phases need to be
on opposite sides of the critical tie line. This is an imaginary tie line
that goes through the Plait point but is parallel to the tie lines in
the two-phase region just before the Plait point is reached.
Through this simple graphical analysis of a gas injection process
it is thus possible to tell a lot about the system, whether there are
opportunities for a vapourising or condensing gas drive and if the
composition of an injected gas could be tailored, by diluting for
example with some amount of the component C2 , so that miscibility
develops. At the same time it is important to keep in mind the
limitations of such a simplistic analysis. Ternary diagrams only apply
at a single pressure and temperature and we have not discussed
at all the significant efforts required through modelling and in the
laboratory to obtain such diagrams, particularly where crude oil is
concerned. The multiple contact process is not a discrete process as
was shown here, rather the gas-rich or oil-rich phases are continuously
evolving as they contact more resident oil or injected gas. Fig. 2.3.9
shows the composition path for a displacement of crude oil by CO2
simulated in a 1D displacement scenario (Gardner et al., 1981). Our
analysis was 0D and did not consider how fluids flow and contact
each other, nor any time constraints on the processes of chemical
71 Vol % UL 73 Vol % UL
87.8 100.0 100 Vol % CO2

Vol % Vol %
Contacts
Forward
UL UL Single Contact Points

12.2 Vol %
29 Vol % Crude 27 Vol % Crude

54.5 Vol % 6.5 Vol % 71.2 Vol %
50 Vol % CO2 CO2 CO2
CO2
54.5 65.0 71.2 73.0

Vol % Vol % Vol % Vol % Plait
UL UL UL UL Point
Swept Zone
Contacts
Li f
Ti n o
ne
al io
ic s
e
45.5
rit en
Vol % 35.5 LL
C Ext
Vol % 28.8 27.0

LL Vol % Vol % 100 100
LL Crude
LL LL Vol %
Vol %
50 Vol % C+2 C–6
Crude 45.5 Vol % 35.0 Vol % 28.8 Vol %
LL LL LL
Figure 2.3.8. A multi-contact mixing experiment performed with Wasson crude

oil and CO2 at 2,000 psi and 105◦ F.
Source: Gardner et al. (1981).
100 Vol % CO2
‘High’ Dispersion
Level Low (Experimental)
Dispersion Level
Plait Point
100 Crude 100

Vol % Vol %
C+2 C –6
Figure 2.3.9. A two-phase ternary diagram from Gardner et al. (1981) showing
the simulated composition path in a 1D displacement for the process shown in
Fig. 2.3.8.
equilibrium. These were thus flash calculations in which we assumed

that equilibrium was instantly reached and that evolving gas rich
fluids were immediately plucked from the mixtures and moved on to
contact new oil once the equilibrium has been (instantly) achieved.
There is some basis for this as the gas rich fluids have higher
mobility than the oil rich fluids. Thus, what is actually observed
in a laboratory core flood or field scale process is described only
qualitatively by this analysis. In practice, for example, the mixing
line for the injection of CO2 into a hydrocarbon needs to be well
above the Plait point in order for miscibility to develop in a reservoir
setting. Perhaps most important, this analysis tells us nothing about
how the fluids themselves will flow in such a process. How much oil
will be recovered and how quickly? How much solvent will be required
to recover a particular amount of oil? Some of these questions will
now be addressed in the next section.
2.3.2. Transport in Two-Phase Multi-Component

Systems
I provide a brief review here of the Buckley–Leverett solution to
immiscible displacement, which we will then extend to incorporate
Plait
point
Critical
tie line
r
r
C C
Figure 2.3.10. A two-phase ternary diagram to be used for exercises.
multi-component fluids with miscibility (Fig. 2.3.10). The following

presentation has been derived primarily from (Pope, 1980; Lake,
1989; Orr, 2007; Blunt, 2013).
2.3.2.1. Review of the Buckley–Leverett solution

The Buckley–Leverett equations provide a solution to a problem of
the displacement of one fluid by another immiscible fluid in one
dimension. In this system, there is no chemical mass transfer between
the phases and thus each fluid phase can be thought of to exist
entirely as a single component. Using fractional flow theory the
governing mass balance expressed in dimensionless form is given by
∂Si ∂fi
+ = 0, (2.3.5)
∂tD ∂xD
where Si is the saturation of a phase i. The fractional flow, fi , of

phase i is the ratio of the flow rate of that phase, qi , to the sum total

of all fluid flow rates, qt = i qi ,
qi
fi = .
qt
Dimensionless time tD and distance xD can be defined by,

qt t
tD = ,
φL
x
xD = ,
L
where L is a reference length scale and φ is the porosity of the system.
In fractional flow theory, capillary pressure gradients are usually
assumed to contribute little to the flow. If we also ignore gravity (i.e.
consider flow in the horizontal direction) fi can be expressed as,
λi
fi = ,
i λi
where the mobility of a phase, λi is the ratio of the relative

permeability of that phase at the phase saturation, kr,i (Si ) to the
viscosity of the phase, µi ,
kr (Si )
λi = .
µi
Thus the fractional flow is a function of saturation and Eq. (2.3.5)

can be written as
∂Si dfi ∂Si
+ = 0. (2.3.6)
∂tD dSi ∂xD
This type of equation lends itself to solution by the method of
characteristics. Using this technique, solutions of the form Si (xD , tD )
are not obtained but rather characteristic curves along which Si is a
constant value, e.g. xD = f (tD )|Si =So . The solution is
dfi
xD = (Si )tD , (2.3.7)
dSi
which says that a saturation, Si , will propagate forward in the
medium at a rate given by the derivative of the fractional flow
function, dfi /dSi , evaluated at that saturation. The derivative of
the fractional flow function is thus referred to as the dimensionless
velocity of the saturation,

dfi
vD = .
dSi
The derivative of the fractional flow curve, however, is often a
non-monotonic function in which case the saturation associated
with a given propagation velocity is non-unique. From a physical
standpoint this is impossible and results in the formation of a shock.
In this case, Eq. (2.3.7) provides the saturation velocities for all
of the saturations greater than the shock saturation, which I will
denote as S ∗ . Saturations below S ∗ have velocities equal to the shock
saturation. The shock saturation and velocity are obtained through
mass balance considerations and given by the equation

∗ df
f1 − f1o
vD = S =S ∗ = , (2.3.8)
dS1
1 S1 − S1o
S1 =S ∗
where S1o and f1o are the saturation and corresponding fractional
flow of phase 1 at an initial condition prior to displacement. In a
graphical representation, S ∗ is the saturation at which the slope of
the fractional flow curve is equal to the slope of the straight line
passing through the points (S1o , f1o ) and [S1∗ , f1 (S1∗ )].
2.3.2.2. Two-component, two-phase displacement

Consider now a system with multiple phases and multiple chemical
components that can move between the phases such that the
component concentration of a chemical i in phase j can be given
by cij . In Sec. 2.3.1, we considered systems of two phases, j = 1, 2,
and three chemical components i = 1, 2, 3.
The total concentration of the component, Ci , will be given by
the sum of the product of the fraction of that component in each
phase with the saturation of those phases,

Ci = cij Sj . (2.3.9)
j
Similarly, the total fractional flux of a chemical component i is given

by the sum of the products of the concentrations in each phase with
the fractional flow of that phase,

Fi = cij fj . (2.3.10)
j
The rate of change of mass of a chemical component per unit volume

is given by the sum of the rates of change in each of the phases,
∂
φ ci,j Sj (2.3.11)
∂t
j
∂Ci
=φ. (2.3.12)
∂t
This must be balanced by fluxes of the component into and out of
the volume. In this class we will consider a 1D system where the net
flux is given by,
∂
ci,j qj (2.3.13)
∂x
j
∂
= ci,j fj qt (2.3.14)
∂x
j
∂Fi
= qt , (2.3.15)
∂x
for incompressible fluid systems where ∂q t
∂x = 0. Note that volume is
often not conserved when components transfer phases, i.e. the density
of CO2 in a hydrocarbon rich phase may well be very different than
the density of CO2 in the CO2 rich phase, and so this is not always
appropriate.
The conservation equation is thus,
∂Ci ∂Fi
+ qt
φ = 0. (2.3.16)
∂t ∂x
We can convert this to dimensionless units using the chain rule,
∂Ci ∂Ci ∂tD qt ∂Ci
= = , (2.3.17)
∂t ∂tD ∂t φL ∂tD
∂Fi ∂Fi ∂xD 1 ∂Fi
= = . (2.3.18)
∂x ∂xD ∂x L ∂xD
The dimensionless mass balance for the chemical component in 1D

is given by
∂Ci ∂Fi
+ = 0. (2.3.19)
∂tD ∂xd
Consider now a two-phase, two-component system, i = 1, 2 and
j = 1, 2. Assuming a two component system is like assuming that
our composition path is constrained to a single tie-line. In a two-
component system, as on a specific tie line, the saturation of a phase,
Sj , and the chemical component concentration in that phase, cij ,
in this system are functions of the total chemical concentration, Ci
only. The phase concentrations are given by the ends of the tie line
associated with Ci and the saturation is determined by the location
of Ci on the tie line, e.g. Eq. (2.3.4). As a result,

Fi = cij fj = cij (Ci )fj (Sj ) = Fi (Ci ), (2.3.20)
j j
and Eq. (2.3.19) can be written as

∂Ci dFi ∂Ci
+ = 0. (2.3.21)
∂tD dCi ∂xd
This is the same type of equation as Eq. (2.3.6), describing the flow
of the fluid phases in the Buckley–Leverett formulation, and thus
can be solved using the method of characteristics for the velocity at
which a total concentration wave propagates in a two-phase system,
dFi
xD = tD . (2.3.22)
dCi
The dimensionless component concentration velocity and the shock
velocities are again given as the derivative of the fraction flow
function,
dFi
vD = , (2.3.23)
dCi
∗ Fi − Fio
vD = . (2.3.24)
Ci − Cio
Component fractional flow curves (plots of Fi versus Ci ) can

be constructed using data describing the phase saturation as a
function of chemical composition, e.g. Eq. (2.3.4), and the relative
permeability curves of the two phases. We will not take this any
further here, however. We will use some of this analysis in a simple
example analysing the properties of first contact miscible water
alternating gas displacement below. The key idea to see is that total
chemical component concentration waves have velocities determined
by a component fractional flow curve, Fi (Ci ), which are related to but
distinct from fluid phase saturation wave velocities that are deter-
mined by the fluid phase fractional flow curve, fj (Sj ). This concept is
general to systems of multiple components whether those components
are in the oil phase (as here) or in the water phase (as in chemical
flooding).
2.3.3. Fractional Flow Theory and Water

Alternating Gas Injection
One of the simplest examples of a multi-component and multi-
phase displacement process is that of a first contact miscible WAG
injection, where a solvent gas, first contact miscible with oil in
the reservoir, is coinjected or injected in alternating slugs with an
aqueous phase, into a reservoir that contains a mixture of an aqueous
phase and oil. This is done with the aim of capturing the benefits
of miscible gas injection, increasing the microscopic displacement
efficiency to 100%, while mitigating the impact of the generally high
mobility of solvent gases on fingering and gravity segregation through
a reduction in the relative permeability.
Consider a 1D displacement wherein water and a hydrocarbon
phase bearing a solvent (the “gas”) are injected at a fractional flow
given by fwo into a reservoir with an initial water and oil saturation at
the connate water saturation given by Sw o . There are only two phases
in this system, a hydrocarbon bearing phase and the aqueous phase

which I will denote with subscripts o and w respectively. Additionally,
there is no chemical component exchange between the two phases and
thus the saturation of the phases is not controlled directly by the
fw
.
water–oil
. water–solvent
. . . .
Sw
Figure 2.3.11. Fractional flow curves for an oil phase displacing water where the
water–solvent curve has a higher mobility ratio.
overall chemical composition of the mixture. The speed at which the

saturation waves propagate is determined by the relevant fractional
flow curves. The injected oil phase with solvent will have a lower
viscosity and there are two fractional flow curves for determining
saturation wave velocities, the curve corresponding to the original
oil–water system and the curve corresponding to the solvent oil–
water system. These are shown in Fig. 2.3.11.
Thus in this system there are two phases plus two components
in the oil phase, the solvent chemical C1 and a pseudo component
consisting of the remainder of the oil. The aqueous phase will be
assumed to consist of a single component that has no solubility in
the oil phase. Thus the total concentration of the solvent, C1 , is
given by,
C1 = c1w Sw + c1o So = c1o So = c1o (1 − Sw ). (2.3.25)
Similarly, the total fractional flux of the solvent, F1 , is given by
F1 = c1w fw + c1o fo = c1o (1 − fw ). (2.3.26)

As a result, the mass balance Eq. (2.3.19) becomes

∂[c1o (1 − Sw )] ∂[c1o (1 − fw )]
+ =0
∂tD ∂xd
∂c1o ∂Sw ∂c1o ∂fw
= (1 − Sw ) − c1o + (1 − fw ) − c1o
∂tD ∂tD ∂xD ∂xD
∂c1o 1 − fw ∂c1o
= + = 0.
∂tD 1 − Sw ∂xD
The dimensionless velocity of the solvent, vD,s , is given by,
1 − fw fo
vD,s = = . (2.3.27)
1 − Sw So
This is seen to be correct when considering that the solvent does
not undergo any mass exchange with the water and tracking the
solvent is akin to tracking a chemical component in a single phase.
In this circumstance the dimensionless velocity is simply given by
the fractional flow divided by the saturation.
The water front velocity is given by the same considerations as
with the Buckley–Leverett formulation, except in this case, fw = 1.
If the fractional flow is such that there is no shock and rarefaction,
the velocity is given by,
fw
vD,w = . (2.3.28)
Sw − Sw
o
In the following discussion we will only consider situations where

saturation changes occur entirely within a shock front.
These equations lend themselves to a graphical characterisation
of the system on plots of the fractional flow versus saturation for
the various fluid pairs. In the following discussion we will use such
graphs to determine the velocities of (1) water shock fronts where
the saturation changes and (2) solvent fronts where the system
changes from an oil–water displacement system to a oil/gas–water
displacement system. Sometimes water shock fronts will occur at
locations distinct from the arrival of the injected gas and many
times a shock front will also be associated with the arrival of the
injected gas.
Figure 2.3.12. Showing the solvent and water injection velocities at an injection
fractional flow where the solvent velocity will be larger than the injected water
velocity
Consider a fractional flow fwo shown in Fig. 2.3.12. The water

saturation associated with this fractional flow is shown as a solid
point in the graph. In this case, very little water is injected relative
to the solvent mixture. The solvent velocity, Eq. (2.3.27), is normally
given by the slope of the line from the top right of the graph
to the saturation point corresponding to the injection fractional
flow. The slope of such a line would be steeper than the slope
of the corresponding line from the point (Sw o , 0) to (s o
w,f , fw ), or
alternatively, the slope of the line from the initial saturation to
the location on the oil–water fractional flow curve, either of which
would indicate the velocity of the associated water shock front. This
suggests that the injected gas moves faster than the water shock
front. Moreover, in this particular case, the gas–oil system is moving
into a resident oil system that is already at the maximum water
saturation. Thus, the arrival of the gas–oil front will not result in a
change in water saturation which is already at the minimum (recall
that the initial condition of the reservoir was at the connate water
saturation). Additionally, the velocity of this front is not given by
Eq. (2.3.27) but rather by the value 1/Sw o . Eventually, a change in
water saturation occurs corresponding to the equilibrium associated

with the fractional flow. The velocity of this is given by Eq. (2.3.28),
the slope of the line from the point on the fractional flow curve at
fractional flow where the solvent velocity will be less than the injected water
velocity.
the initial water saturation (in this case fw = 0 at this point) to

the saturation point on the solvent–water fractional flow curve. The
velocity of the solvent in this region is given by Eq. (2.3.27), but this
is of less importance.
Now consider the system with an injection ratio with a high
fraction of water, fwo shown in Fig. 2.3.13. The gas velocity is
given by Eq. (2.3.27) but it is apparent that the water saturation
wave will move faster than the solvent. Thus, initially, the water
saturation will increase even though the gas has not yet arrived.
An increase in water saturation in the absence of the arrival of the
gas indicates that this wave velocity will be determined by the oil–
water fractional flow curve. Thus, there will be some saturation,
Swf , on the oil–water fractional flow curve (Swf ,fwf ) associated
with the wave that arrives prior to the injected gas. This saturation
will not necessarily be equivalent, however, to the final saturation,
Ssf , on the solvent–water fractional flow curve, determined by the
fractional flow of the injection itself. This intermediary saturation is
found by mass balance considerations. The gas (i.e. solvent) velocity
at the final saturation, Ssf , must be equivalent to the oil velocity
at the intermediary saturation, Swf . Formally, the line defined by
Eq. (2.3.27) is extended to the oil–water fractional flow curve. Their
fractional flow where the solvent velocity is equal to the injected water velocity.
point of intersection determines the saturation of the leading water

shock front,
1 − fwo f o − fwf
vD,s = = w . (2.3.29)
1 − Ssf Ssf − Swf
The velocity of the leading shock front is given by the slope of the
line from the initial saturation to (Swf ,fwf ). Again, in this case the
saturation change is entirely within the shock front and there is no
subsequent spreading wave.
Finally, the system with injection ratio fwo shown in Fig. 2.3.14
is the so-called optimum WAG ratio because it is the fractional flow
for which the water and solvent fronts will move at the same velocity.
2.3.4. Three-Phase Flow

There are many cases in EOR where flow will involve three phases, oil
gas and water. Thus extensions of the flow analysis presented above
to three-phase flow is also useful.
2.3.4.1. Fractional flow theory applied to three-phase flow

Consider simultaneous flow of an oil, gas and aqueous phase that will
be denoted with subscripts o, g and w, respectively. The saturation
of the phases are related by,
So = 1 − Sw − Sg . (2.3.30)
The definitions of the fractional flow of each phase in the absence of

capillary pressure gradients and gravity are given by,
λw
fw = , (2.3.31)
λt
λg
fg = . (2.3.32)
λt
The mass balance equations in terms of the fractional flow of each
phase can be written as
∂Sw ∂fw
+ = 0, (2.3.33)
∂tD ∂xD
∂Sg ∂fg
+ = 0. (2.3.34)
∂tD ∂xD
The fractional flows, however, are functions of two variables (Sg and
Sw ) and thus expansion of Eq. (2.3.33) and (2.3.34) is given by,
∂Sw ∂fw ∂Sw ∂fw ∂Sg
+ + = 0, (2.3.35)
∂tD ∂Sw ∂xD ∂Sg ∂xD
∂Sg ∂fg ∂Sw ∂fg ∂Sg
+ + = 0. (2.3.36)
∂tD ∂Sw ∂xD ∂Sg ∂xD
Assuming that there are specific wave speeds, vD , associated with sat-
urations, as with the two-phase displacement of the Buckley–Leverett
approach, results in the quadratic partial differential equation

∂fw ∂fg ∂fw ∂fg
vD − vD − − = 0. (2.3.37)
∂Sw ∂Sg ∂Sg ∂Sw
For a given water and gas saturation, there are two values of vD that
satisfy Eq. (2.3.37).
2.3.4.2. Models for three-phase relative permeability

As with two-phase flow, the description of three-phase flow is
sensitive to the relative permeability curves through the fractional
flow functions. In the case of three-phase flow, however, relative
permeability is not well understood from a theoretical standpoint and
there is generally very little experimental data from which to draw
empirical correlations. Summaries of the major issues are included
in (Blunt, 2000; Juanes and Patzek, 2004; Baker, 1988; Delshad and
Pope, 1989). This is an area of ongoing research but in summary,
models currently considered the most accurate for representing three-
phase relative permeability interpolate between values of two-phase
relative permeabilities. In other words, for the oil relative perme-
ability an interpolation is made between the relative permeability in
the oil–water system, kr,o(w) and in the oil–gas system, kr,o(g) , with
the interpolation weighted by the water and gas saturations. The
equivalent can be done for the water and gas relative permeabilities
(Blunt, 2000). Following the model proposed by (Baker, 1988), the
relative permeabilities for oil, water and gas can be given as
(Sw − Sw,i )kr,o(w) + (Sg − Sg,r )kr,o(g)
kr,o = , (2.3.38)
(Sw − Sw,i ) + (Sg − Sg,r )
(So − So,i )kr,w(o) + (Sg − Sg,r )kr,w(g)
kr,w = , (2.3.39)
(So − So,i ) + (Sg − Sg,r )
(Sw − Sw,i )kr,g(w) + (So − So,i )kr,g(o)
kr,g = . (2.3.40)
(Sw − Sw,i ) + (So − So,i )
In this model, Sw,i is the irreducible water saturation, Sg,r is the
residual gas saturation in an oil–water displacement and So,i is the
initial oil saturation in a gas–water displacement, the latter two are
usually zero. Using this model, the three-phase relative permeability
functions are obtained through the measurement of six two-phase
relative permeability curves, both phases in an oil–water, oil–gas and
gas–water displacement.
2.4. Chemical EOR: Polymers and Low

Salinity Flooding
Chemical EOR techniques involve changing the chemistry of the
injection water either by adding chemical components (polymers,
Alkali–Surfactant–Polymer (ASP), sulphate salts) or removing them
(low salinity water flooding) in order to improve oil recovery. Of these
techniques, polymer flooding has the longest history (dating back to
the late 1960s) and is the only one to have been deployed at scale
in the field. ASP has been discussed for almost as long as polymer
flooding although it has only been used in field pilots to date. It was
devised to combine the benefits of using a surfactant to reduce the
oil–water interfacial tension and hence reduce capillary trapping of
the oil and a polymer to provide mobility control whilst including an
alkali to prevent both of these being lost by adsorption onto the rock.
In contrast, low salinity water injection has only been extensively
discussed since it was rediscovered by (Yildiz and Morrow, 1996).
There have been several single well field tests as well as one interwell
field trial (Seccombe et al., 2010) but it will not be deliberately
deployed until Clair Ridge comes on stream in 2017. High sulphate
water injection is a specialised EOR technique for chalk reservoirs
e.g. (Zhang et al., 2007). Only polymer flooding and low salinity
water flooding will be discussed in detail in this section.
2.4.1. Polymer Flooding

In polymer flooding, water soluble polymers are added to the
injection water to viscosify the water. It is usually used for the
recovery of more viscous, but still mobile oils (up to a viscosity of
approximately 0.1 Pa·s) in higher permeability reservoirs (several
hundred mD or higher). Using a polymer solution increases the
mobility ratio and thus reduces the tendency for the water to bypass
the oil through either viscous fingering or channelling. Additional
oil may also be recovered by increased viscous cross-flow (Sorbie,
1991) thus polymer flooding is the only one of the EOR processes
targeted at improving macroscopic sweep as well as improving
microscopic displacement efficiency. One of the main disadvantages of
this process, apart from cost, is that injection rates tend to be lower
because of the increased water viscosity, thus overall production rate
may be reduced. Nonetheless, oil production rates may be higher
because of the reduced water cut. Some polymers are shear thinning,
meaning that their viscosity reduces at high shear rates (such as
those that occur in and near the wellbore). In this case, there will be
a smaller reduction in injectivity. Further problems may arise if the
polymer is adsorbed onto the rock or gets trapped in the pore throats.
The first effect will tend to reduce the effectiveness of the polymer
(as there will be less in solution and thus the solution viscosity will
not be as high as planned) whilst the second effect may result in

permeability reduction and further loss of injectivity.
The principal chemicals used are polyacrylamide (used in 95%
of projects in the literature) and xanthan gum e.g. (Morel et al.,
2008). Polyacrylamide is the cheaper of the two and does not degrade
easily. This latter attribute also means that it persists in the natural
environment and so it cannot be applied in Norwegian oil fields
unless the produced water is reinjected. It is also less stable in
higher salinity reservoir waters. Xanthan gum is more stable at these
higher salinities but degrades too easily for most reservoir uses. It
is also more expensive. Neither of these chemicals is stable at very
high temperatures (i.e. deep reservoirs) or in carbonate reservoirs
where the connate water is particularly hard. Current research is
focussed on developing more robust, cheap and environmentally
friendly alternatives.
At the time of writing, China is the only country in the world
where there has been and continues to be large scale oil produc-
tion from polymer flooding. This is primarily a result of project
economics. Field scale deployment of polymer flooding needs large
volumes of chemicals (e.g. using xanthan in the proposed Dalia devel-
opment, offshore Angola, would have required a 50% increase in world
production of xanthan, (Morel et al., 2008)), transportation of those
chemicals and then the facilities to mix and inject those chemicals
without breaking down the molecules and thus reducing their viscosi-
fying effect. Nonetheless in 2004 it was reported that 250,000 barrels
of oil a day were being produced as a result of polymer injection into
the Daqing field in China (Chang et al., 2006). The high oil price until
2014 combined with the increased focus on maintaining production
from mature fields led to renewed interest in polymer flooding and
tests have begun on commercial scales using both polymer floods and
ASP flooding in Oman (Stoll et al., 2011; Al-Saadi et al., 2012). It
has also been identified as one of the three highest potential EOR
technologies for the North Sea (McCormack et al., 2014).
Fractional flow analysis can be used to understand how polymer
flooding increases the microscopic displacement efficiency. This
analysis is similar to the miscible gas injection process described
above with two key differences: first, the dissolved chemical con-
centration, C1 , of interest is polymer and it is dissolved in the
aqueous phase. Second, one of the key processes to capture in
polymer flooding is the reversible adsorption of the polymer onto
solid walls of the rock pores. This is done by accounting for some
amount of the concentration, C1 , as made up of the adsorbed
concentration, c1a , the amount of adsorbed polymer per unit pore
volume. Figure 2.4.1 shows two fractional flow curves representing
the displacement processes between the water–oil system and the
aqueous polymer solution–oil system, with the mobility lower in
the polymer solution displacement due to the increased viscosity.
In this iteration thus the total concentration, including the adsorbed
component, is given by,
C1 = c1w Sw + c1o So + c1a = c1w Sw + c1a .
The total flux is given by,
F1 = c1w fw + c1o fo = c1w fw .
It is usually assumed that there exists a simple relationship between

the concentration of polymer in the aqueous phase and the adsorbed
concentration, e.g.
c1a = Rc1w ,
where R is a proportionality constant known as the retardation

factor. This is because processes of reversible adsorption have the
Figure 2.4.1. Fractional flow analysis for polymer flooding.

ultimate effect of slowing down a concentration wave as will be seen

below.
From this, the mass balance Eq. (2.3.19) becomes
∂[c1w Sw + c1a ] ∂[c1w (fw )]
+ =0
∂tD ∂xd
∂c1w ∂Sw ∂c1a ∂c1w ∂fw
= Sw + c1w + + fw + c1w
∂tD ∂tD ∂tD ∂xD ∂xD
∂c1w ∂c1a ∂c1w ∂c1w
= Sw + + fw
∂tD ∂c1w ∂tD ∂xD
∂c1w ∂c1w
= (Sw + R) + fw
∂tD ∂xD
∂c1w fw ∂c1w
= + = 0.
∂tD Sw + R ∂xD
The velocity of the concentration of the polymer solution is thus
fw
vsw = .
Sw + R
Consider a polymer flood where the aqueous polymer solution is
injected at a fractional flow of 1 (Fig. 2.4.1). There will be two shock
fronts: A leading shock front where resident brine is banked and a
second shock front between the polymer solution and the resident
brine. The particular saturation, fractional flow and velocities of
those fronts are found through mass balance considerations. In
particular, the shock saturation for the arrival of the polymer food
is given by equating,

fw dfw
= .
Sw,3 + R dSw
Sw =Sw,3
Similarly, the saturation front of the resident water bank is given

by the following equation further to the concentration shock velocity
between the polymer solution and the resident brine,
fw,3 fw,3 − fw,2
= .
Sw,3 + R Sw,3 − Sw,2
Finally there is a rarefaction of the saturation wave to the final
residual oil saturation.
2.4.2. Low Salinity Water Flooding

Low salinity water flooding involves the injection of water with
a reduced salt concentration. It appears to work by making the
reservoir rock more water-wet, reducing the residual oil saturation
and changing the fractional flow behaviour. It is therefore targeted
at improving microscopic displacement efficiency. In principal, it
is one of the cheaper and simpler chemical EOR techniques as it
only needs the installation of desalination plants rather than the
injection of large volumes of expensive chemicals. There is still some
debate about the exact mechanism causing this wettability alteration
(e.g. Lager et al., 2008; Tang and Morrow, 1999) and indeed some
authors (Thyne, 2011; Skrettingland et al., 2011) question whether
there is any improved oil recovery at all, resulting from this process.
Nonetheless, most of the literature suggests that there will be
improved oil recovery (e.g. Webb et al., 2003; Mahani et al., 2011).
There appear to be four criteria necessary for low salinity water-
flooding to be a success: (a) the crude oil should contain a significant
proportion of polar components (especially organic acids), (b) the
connate water should contain divalent cations (especially Ca2+ and
Mg2+ ), (c) the reservoir rock must be sandstone, ideally with a
proportion of clays, especially kaolinite and (d) the injected water
must be softened, i.e. has a reduced divalent cation content. These
criteria are most consistent with the multi-component ion exchange
mechanism for low salinity waterflooding EOR as originally proposed
by (Lager et al., 2008). More recent research (e.g. Hassenkam et al.,
2011) also tends to support this model but it is possible that the
release and migration of clay fines may also play a role in some rocks
(Tang and Morrow, 1999).
To date there have been no deliberate field wide implementations
of low salinity water injection although there have been some
inadvertent deployments (e.g. in the Omar field in Syria, (Mahani
et al., 2011)). There have been a number of single well tracer tests
and one well-to-well pilot (Webb et al., 2003). The first planned
low salinity development will be at Clair Ridge in the North Sea
with production due to start in 2017 (McCormack et al., 2014) and,
along with polymer flooding and gas injection, it is one of the three
EOR techniques with most potential for improving oil recovery in

the North Sea as identified by (McCormack et al., 2014). The main
practical factor limiting its deployment is the availability of space for
the desalination plants on ageing platforms.
The simplest model for predicting the performance of low salinity
waterfloods is that of (Jerauld et al., 2008). In this model, there
are two sets of relative permeability curves, a more oil wet pair of
curves corresponding to a high salinity water-flood and a more water-
wet pair, corresponding to the low salinity waterflood. The reservoir
originally contains oil and high salinity connate water. Lower salinity
water is injected displacing both the high salinity water and some of
the oil. This results in a 1D displacement formed of two shocks and a
rarefaction, similar to that seen when performing the fractional flow
analysis for polymer flooding shown in the previous section. As in
the polymer flooding case there is a connate water bank ahead of the
low salinity front.
As in the polymer flooding case we draw a graph showing the
fractional flow curve for the high salinity water injection relative
permeabilities and the fractional flow curve resulting from using the
low salinity water injection relative permeabilities. In this case, we
draw a line from the point (0, 0) (see Fig. 2.4.2) that is tangent to
the low salinity fractional flow curve. This is because the low salinity
and high salinity water are fully miscible and we assume there is
no retardation of the low salinity water. The point of intersection
Figure 2.4.2. Fractional flow analysis for low salinity water flooding applied for
primary recovery.
of this line with the high salinity fractional flow curve (Sw,2 , fw,2 )
gives the shock front saturation of the leading edge of the connate
water bank (Sw,2 ) and the point of tangency gives the shock front
saturation of the leading edge of the low salinity water (Sw,3 ). The
characteristic velocity of the leading shock is determined by drawing
a line from the connate water saturation of the high salinity fractional
flow to the point (Sw,2 , fw,2 ) and calculating the gradient of that
line. The characteristic velocity of the trailing shock is simply the
gradient of the original tangent line. Finally, there is a rarefaction
to the residual oil saturation. If low salinity flooding is applied as
a secondary or tertiary recovery process, the oil saturation will be
low. This will result in the creation of an oil bank, characteristic of
enhanced recovery induced by a wetting state change (Fig. 2.4.3).
Jerauld et al. (2008) proposed a refinement of this model when
implementing it in a numerical simulator. They asserted that the
injected low salinity water would only begin to alter oil recovery
when the water salinity was lower than a specified upper salinity
threshold (e.g. 7,000 ppm) and that it would only be fully effective
once salinity was below a lower threshold (e.g. 1,000 ppm). For
salinities below the lower threshold then the flow would be described
by the more water-wet relative permeabilities whilst for salinities
above the upper threshold the flow would be described by the more
oil-wet relative permeabilities. For intermediate salinities then the
relative permeabilities would be a linear combination of the high
Figure 2.4.3. Fractional flow analysis for low salinity flooding applied as a
secondary or tertiary recovery mechanism.
and low salinity relative permeabilities. This allows the approximate

modelling of the effect of mixing between the injected low salinity
water and the high salinity connate water on oil recovery. More
sophisticated models (such as those of Dang et al. (2013)) use a
more sophisticated ion exchange model to determine which relative
permeability curves should be used during the displacement.
2.5. Practical Considerations

In Sec. 2.1, we saw that EOR contributed only 2% to global oil
production in 2014. Current predictions suggest that it will still
contribute less than 10% by 2030 (Aleklett et al., 2010). We have
seen that the average recovery factor from oil fields around the world
is between 30% and 40% although in some fields such as Forties in the
North Sea it is believed the recovery factor has reached 65%. We also
know that it is becoming increasingly difficult to find new oil fields
that contain significant volumes of conventional oil. Instead the world
is looking increasingly towards the exploitation of unconventional oil
and gas reserves such as so called shale oil and gas and the very
viscous bitumens found in Canada and Venezuela. Given the variety
of EOR techniques available to reservoir engineers and the fact that
the majority of these have been discussed, researched and often
piloted in the field for more than 40 years, why is the contribution
of EOR to world production so low and projected to grow so little?
The reality is that, although it is technically feasible to signif-
icantly improve oil recovery using EOR techniques, it is often not
economically feasible to do so. EOR requires very large volumes of
gas or chemicals. These injectants are more expensive than water and
the benefits of using them may not be seen for many years. This is one
reason why national oil companies (e.g. in China) are more likely to
implement EOR than independent oil companies. Governments may
take a longer term view, husbanding their nation’s resources for the
benefit of the country whilst independent oil companies need to pay
dividends to their shareholders every year.
Sometimes there are practical considerations limiting the deploy-
ment of EOR e.g. there is simply not enough hydrocarbon gas
available to replace the produced oil and maintain reservoir pressure

above the minimum miscibility pressure and it is not economic to
build pipelines to import that gas from elsewhere. We also know that
hydrocarbon gas in itself is a valuable commodity. There is therefore
a commercial decision to be made balancing the immediate income
from gas sales versus the additional oil that may result some years
after injecting the gas into the reservoir. Most successful miscible gas
injection projects have been in fields producing significant quantities
of gas that cannot easily be sold because the field is very remote
from centres of population that could use the gas (e.g. Prudhoe Bay
in Alaska) or because there is no export route for the produced gas
(e.g. Ula in the North Sea).
From the chemical perspective we have seen, for example, that
using xanthan as the polymer in the Dalia field would have required
a 50% increase in world production of this chemical (Morel et al.,
2008). In general it is not economic to build whole new factories
to supply this volume of polymer. In addition, many oil fields are in
geographically remote regions of the world, often in extreme climates
(e.g. the Arctic, or the desert). In the Arctic, it may only be possible
to transport materials to the well head for the 6 months of the year
when the ground is frozen. Transporting large volumes of polymer
(even if they were available) to the middle of the desert may require so
many trucks that it may not be practical. Even if it were, there may
not be sufficient good quality water available in which to dissolve
it prior to injection. This is also a factor that limits the potential
deployment of low salinity water injection.
It is, of course, also important to bear in mind environmental
considerations when planning an EOR scheme. The chief concern
in most schemes is the disposal of produced water. Even ordinary
produced water needs to be reinjected or treated before it can be
disposed of on the surface. This treatment becomes more important
when chemicals, such as polymers or surfactants, are injected in
the solution. We have already noted that polyacrylamide cannot be
injected into Norwegian fields because it persists in the environment.
Most, if not all, EOR techniques are most effective if applied as
secondary recovery schemes. This is because water will tend to trap
oil in pores and bypass oil in lower permeability zones. EOR injectant
applied after a waterflood may then simply follow the flow paths
already established by the injected water so less oil will be accessed.
It will also take longer for the oil that is swept up by the EOR
injectant to reach the production well. Nonetheless, many engineers
will only consider using EOR after a waterflood. This may be because
at the time the field was discovered the oil price was too low to
justify the additional expense of an EOR scheme. However, even
when the oil price has been higher many engineers have preferred
to start production using water injection. In this case, they reason
that this will give them a better understanding of the geological
heterogeneity in their reservoir and will thus reduce the risk of a
poor outcome when they finally deploy EOR. It is not always clear
that this reduction in perceived risk outweighs the loss of additional
recovery from applying the EOR scheme in secondary mode. The
response from tertiary EOR schemes may be so slow that it can take
a year or more for there to be a noticeable slowing of the decline in oil
rate or indeed an observable improvement in oil rate. As an example
it took more than two years from the start of gas injection in Magnus
to see a noticeable improvement in oil recovery (Brodie et al., 2012).
It can also be difficult to clearly relate this improvement in oil rate
to the EOR process rather than more recent activities such as well
workovers. In any case, this delay may make the scheme uneconomic.
If it is likely that EOR will be applied post-waterflood then it
is important to plan the field infrastructure to allow for this. If this
is not done then it may become prohibitively expensive to retrofit
these facilities e.g. many offshore platforms on older fields in the
North Sea have no space for the installation of desalination plants
or the preparation of polymer solutions. Injection of carbon dioxide,
whilst potentially effective in terms of recovering additional oil and
also providing a means of mitigating climate change through the
storage of the gas, would require installation of specialised pipework
and wells that are corrosion resistant. To give an indication of the
potential costs, in 2014 drilling a new well in the North Sea could
cost £100 million. In the case of CO2 EOR/sequestration there
would also be the additional cost of a subsea pipeline network to
distribute the CO2 to fields from the power stations where it would be
captured.
Leaving aside economic considerations, the appraisal, planning
and development of EOR schemes can be significantly more complex
than for a conventional water flood. We have already seen in the
previous chapters how 1D fractional flow analysis for WAG, polymer
and low salinity water injection is more involved than for water
flooding, not to mention the additional understanding of phase
behaviour that is needed when planning a miscible gas injection
scheme. Leaving aside the logistical considerations discussed in
preceding paragraphs and the additional facilities design required,
further specialised laboratory tests may be needed to determine
EOR specific data such as the parameters needed to describe phase
behaviour when injecting a miscible or multi-contact miscible gas,
the relative permeability curves needed to plan a low salinity flood
in the reservoir of interest or to quantify the expected levels of
polymer adsorption. Having estimated the microscopic displacement
efficiency of possible EOR schemes using the analytical techniques
described in previous chapters, it will then be necessary to assess
the impact of geological heterogeneity on macroscopic sweep as well
as the range of possible outcomes due to the uncertainty in that
heterogeneity. This requires the use of specialised reservoir simulation
software which in turn relies on specialised reservoir engineering
expertise (e.g. chemistry of rock–fluid interactions for low salinity
waterflooding, chemistry and flow behaviour of the emulsions that
form during ASP flooding, phase behaviour of multi-contact miscible
displacements and use of equations of state). In some cases this
additional complexity may be sufficient to dissuade engineers and
their managers from even considering an EOR scheme.
Despite these various economic, practical and technical diffi-
culties EOR still has the potential to significantly improve oil
recovery. McCormack et al. (2014) suggest that EOR could recover
an additional 0.6–1.2 billion barrels of oil, from the UKCS alone.
Worldwide it could increase the reserves to production ratio from 40
years to perhaps 45 or 50 years. This can only be achieved, however,
through a combination of good reservoir engineering, longer term
planning, innovative research and, of course, a sufficiently high oil

price to make these schemes economic. Burning oil in the form of
petrol and diesel does increase the amount of CO2 in the atmosphere
and is thus one of the cause of anthropogenic climate change, but at
the time of writing this chapter there is no viable alternative fuel for
transportation that can be immediately be applied worldwide in all
forms of transport. Until this becomes available crude oil will still be
needed and EOR has the potential to help supply that demand.
References
Al-Saadi, F. S., Amri, B. A., Nofli, S., Wunnik, J. V., Jaspers, H. F., Harthi,
S., Shuaili, K., Cherukupalli, P. K. and Chakravarthi, R. (2012). Polymer
flooding in a large field in South Oman — initial results and future plans.
SPE EOR Conference at Oil and Gas West Asia held in Muscat, Oman,
(SPE 154665).
Aleklett, K., Hook, M., Jakobsson, K., Lardelli, M., Snowden, S. and
Soderbergh, B. (2010). The peak of the oil age — analyzing the world
oil production reference scenario in World Energy Outlook 2008. Energ.
Policy, 38(3), 1398–1414.
Anderson, W. (1987). Wettability literature survey — part 6: The effects of
wettability on waterflooding. J. Petrol. Technol., 39(12), 1605–1622.
ARI (2009). CO2 storage in depleted oilfields: Global application criteria for
carbon dioxide enhanced oil recovery. Technical Report 2009–12, IEA
Greenhouse Gas Program.
Baker, L. E. (1988). Three phase relative permeability correlations. SPE/DOE
Enhanced Oil Recovery Symposium Tulsa, Oklahoma, (SPE/DOE 17369).
Blunt, M. J. (2000). An empirical model for three-phase relative permeability.
Soc. Petrol. Eng. J., 5(4), 435–445.
Blunt, M. J. (2013). Reservoir Performance Predictors Course Notes. Imperial
College London.
Brodie, J., Jhaveri, B., Moulds, T. and Mellemstrand, H. S. (2012). Review of
gas injection projects in BP. In Proceedings of the 18th SPE Improved Oil
Recovery Symposium, Tulsa, OK 14–18 April.
Chang, H., Zhang, Z., Wang, Q., Zu, Z., Guo, Z., Sun, H., Cao, X. and Qiao,
Q. (2006). Advances in polymer flooding and alkaline/surfactant/polymer
processes as developed and applied in the People’s Republic of China.
J. Petrol. Technol., 58(2), 84–91.
Christie, M., Jones, A. and Muggeridge, A. (1990). Comparison between labora-
tory experiments and detailed simulations of unstable miscible displacement
influenced by gravity. North Sea Oil and Gas Reservoirs-II.
Dang, C., Nghiem, L. X., Chen, Z. and Nguyen, Q. P. (2013). Modeling
low salinity waterflooding: Ion exchange, geochemistry and wettability
alteration. SPE Annual Technical Conference and Exhibition, New Orleans,

Louisiana, USA, 30 September–2 October, (SPE 166447).
Delshad, M. and Pope, G. A. (1989). Comparison of the three-phase oil relative
permeability models. Transport in Porous Media, 4, 59–83.
Dietz, D. (1953). A theoretical approach to the problem of encroaching and
bypassing edge water. Pro. Koninkl. Ned. Akad. Wetemchap., S36.
Djabbarov, S. (2014). Experimental and numerical studies of first contact miscible
injection in a quarter five spot pattern. Master’s thesis, Imperial College
London.
Dumore, J. (1964). Stability considerations in downward miscible displacements.
Soc. Petrol. Eng. J., 4(4), 356–362.
Dykstra, H. and Parsons, R. (1950). The prediction of oil recovery by water-
flood. Secondary Recovery of Oil in the United States, 2nd ed. Dallas: API,
pp. 160–174.
Fayers, F. J. and Muggeridge, A. H. (1990). Extensions to Dietz theory and
behavior of gravity tongues in slightly tilted reservoirs. SPE Reservoir
Eng., 5(04), 487–494.
Gardner, J. W., Orr, F. M. Jr. and Patel, P. (1981). The effect of phase behavior
on CO2 -flood displacement efficiency. J. Petrol. Technol., 33(11), 2061–
2081.
Hassenkam, T., Pedersen, C., Dalby, K., Austad, T. and Stipp, S. (2011).
Pore scale observation of low salinity effects on outcrop and oil reservoir
sandstone. Colloid Surface A, 390, 179–188.
Hite, J., Stosur, G., Carnaham, N. and Miller, K. (2003). Guest editorial. ior
and eor: effective communication requires a definition of terms. J. Petrol.
Technol., 55(16).
Homsy, G. (1987). Viscous fingering in porous media. Annu. Rev. Fluid Mech.,
19, 271–311.
IEA (2013a). Resources to Reserves 2013 — Oil, Gas and Coal Technologies for
the Energy Markets of the Future. International Energy Agency.
IEA (2013b). World Energy Outlook 2013. International Energy Agency.
Jensen, J. and Currie, I. (1990). A new method for estimating the Dykstra-
Parsons coefficient to characterize reservoir heterogeneity. SPE Reservoir
Eng., 5(3), 369–374.
Jerauld, G. R., Lin, C. Y., Webb, K. J. and Seccombe, J. C. (2008). Modeling
low-salinity waterflooding. SPE Reserv. Eval. Eng., 11(6), 1000–1012.
Juanes, R. and Patzek, T. W. (2004). Three-phase displacement theory: An
improved description of relative permeabilities. Soc. Petrol. Eng. J., 9(3),
1–12.
Koottungal, L. (2014). 2014 worldwide EOR survey. Oil and Gas J., 04/07/2014.
Koval, E. (1963). A method for predicting the performance of unstable miscible
displacements in heterogeneous media. Soc. Petrol. Eng. J., 3(02), 145–154.
Lager, A., Webb, K. J., Collins, I. R. and Richmond, D. M. (2008). LoSal
enhanced oil recovery: Evidence of enhanced oil recovery at the reservoir
scale. SPE Symposium on Improved Oil Recovery, Tulsa, Oklahoma, USA,
20–23 April (SPE-113976-MS).
Lake, L. W. (1989). Enhanced Oil Recovery. Society of Petroleum Engineers,

Texas, United States.
Lenormand, R., Zarcone, C. and Sarr, A. (1983). Mechanism of the displacement
of one fluid by another in a network of capillary ducts. J. Fluid Mech., 135,
337–353.
Mahadevan, J., Lake, L. W. and Johns, R. T. (2003). Estimation of true
dispersivity in field-scale permeable media. Soc. Petrol. Eng. J., 8(3), 272–
279.
Mahani, H., Sorop, T. G., Ligthelm, D., Brooks, A. D., Vledder, P., Mozahem, F.
and Ali, Y. (2011). Analysis of field responses to low salinity waterflooding
in secondary and tertiary mode in syria. SPE Europec/EAGE Annual
Conference and Exhibition, Vienna, Austria, (SPE 142960).
McCormack, M. P., Thomas, J. M. and Mackie, K. (2014). Maximising enhanced
oil recovery opportunities in ukcs through collaboration. Abu Dhabi
International Petroleum Exhibition and Conference, (SPE 172017).
Morel, D., Vert, M., Jouenne, S. and Nahas, E. (2008). Polymer injection in
deep offshore field: The dalia case. SPE Annual Technical Conference and
Exhibition, Denver, Colorado, September, (SPE 116672).
Muggeridge, A., Cockin, A., Webb, K., Frampton, H., Collins, I., Moulds, T. and
Salino, P. (2012). Recovery rates, enhanced oil recovery and technological
limits. Philos. T. Roy. Soc. A, 372(20120320).
Orr, F. M. Jr., Yu, A. D. and Lien, C. L. (1981). Phase behavior of co2 and crude
oil in low-temperature reservoirs. Soc. Petrol. Eng. J., 21(4), 480–492.
Orr, F. M. Jr. (2007). Theory of Gas Injection Processes. Tie-Line Publications.
Pentland, C. H., Itsekiri, E., Mansoori, S. K. A. and Iglauer, S. (2010).
Measurement of nonwetting-phase trapping in sandpacks. Soc. Petrol. Eng.
J., 15(02), 274–281.
Pope, G. A. (1980). The application of fractional flow theory to enhanced oil
recovery. Soc. Petrol. Eng. J., 20(03).
Roof, J. (1970). Snap-off of oil droplets in water-wet pores. Soc. Petrol. Eng. J.,
10(01), 85–91.
Salathiel, R. (1973). Oil recovery by surface film drainage in mixed-wettability
rocks. J. Petrol. Technol., 25(10), 1216–1224.
Schmalz, J. P. and Rahme, H. D. (1950). The variation of waterflood performance
with variation in permeability profile. Producers Monthly, 15(9), 9–12.
Seccombe, J., Lager, A., Jerauld, G., Jhaveri, B., Buikema, T., Bassler, S., Denis,
J., Webb, K., Cockin, A. and Fueg, E. (2010). Demonstration of low-salinity
EOR at interwell scale, Endicott field, Alaska. SPE Improved Oil Recovery
Symposium, Tulsa, Oklahoma, 24–28 April, (SPE-129692-MS).
Seright, R. S., Lane, R. H. and Sydansk, R. D. (2003). A strategy for attacking
excess water production. SPE Prod. Facil., 18, 158–169.
Skrettingland, K., Holt, T., Tweheyo, M. T. and Skjevark, I. (2011). Snorre low-
salinity-water injection-coreflooding experiments and single-well field pilot.
SPE Reserv. Eval. Eng., 14(2), 182–192.
Smalley, P. C., Ross, B., Brown, C. E., Moulds, T. P. and Smith, M. J. (2009).
Reservoir technical limits: A framework for maximizing recovery from oil
fields. SPE Reserv. Eval. Eng., 12(04), 610–617.
Sorbie, K. S. (1991). Polymer-Improved Oil Recovery. Springer.
Sorbie, K. S. and Seright, R. S. (1992). Gel placement in heterogeneous systems
with crossflow. Proceedings SPE/DOE Enhanced Oil Recovery Symposium,
Tulsa, OK, (SPE-24192-MS).
Stoll, W., al Shureqi, H., Finol, J., Al-Harthy, S., Oyemade, S., de Kruijif, A.,
van Wunnik, J., Arkesteijn, F., Bouwmeester, R. and Faber, M. (2011).
Alkaline/surfactant/polymer flood: From the laboratory to the field. SPE
Reserv. Eval. Eng., 14(06), 702–712.
Taber, J., Martin, F. and Seright, R. (1997a). Eor screening criteria revisited —
part 1: Introduction to screening criteria and enhanced recovery field
projects. SPE Reservoir Eng., 12(3), 189–198.
Taber, J., Martin, F. and Seright, R. (1997b). EOR screening criteria revisited —
part 2: Applications and impact of oil prices. SPE Reservoir Eng., 12(03),
199–205.
Tang, G. Q. and Morrow, N. R. (1999). Salinity, temperature, oil composition
and oil recovery by waterflooding. SPE Reservoir Eng., 12, 269–276.
Tchelepi, H. A. and Orr, F.M. Jr. (1994). Interaction of viscous fingering,
permeability heterogeneity, and gravity segregation in three dimensions.
SPE Reservoir Eng., 9(04), 266–271.
Thyne, G. (2011). Evaluation of the effect of low salinity waterflooding for 26 fields
in Wyoming. SPE Annual Technical Conference and Exhibition, Denver,
Colorado, (SPE 147410).
Todd, M. and Longstaff, W. (1972). The development, testing and application of
a numerical simulator for predicting miscible flood performance. J. Petrol.
Technol., 24(7), 874–882.
USEIA (2014). Annual Energy Outlook 2014 with projections to 2014. Technical
Report DOE/EIA-0383(2014), US Energy Information Administration.
Wallace, M. and Kuuskraa, V. (2014). Near-term projections of CO2 utilization
for enhanced oil recovery. Technical Report DOE/NETL-2014/1648,
National Energy Technology Laboratory.
Webb, K. J., Black, C. J. J. and Al-Ajeel, H. (2003). Low salinity oil recovery- log
inject log. Middle East Oil Show, Bahrain, 9–12 June, (SPE-81460-MS).
Yildiz, H. O. and Morrow, N. R. (1996). Effect of brine composition on recovery
of Moutray crude oil by waterflooding. J. Petrol. Sci. Eng., 14, 159–168.
Zhang, P. M., Tweheyo, M. T. and Austad, T. (2007). Wettability alteration and
improved oil recovery by spontaneous imbibition of seawater into chalk:
impact of the potential determining ions Ca2+ , Mg2+ and SO−4 2 . Colloid
Surface A, 301, 199–208.
Chapter 3
Numerical Simulation
Dave Waldren
Petroleum Consulting and Training Ltd.
Cobham, Surrey, UK
3.1. Reservoir Models

3.1.1. Introduction
The purpose of modelling, in any discipline, is to be able to simulate
a process or physical phenomenon in a controlled environment in
order to monitor the response of the model to a coherent set of
environmental changes.
Models may be broadly classified into the categories of
(1) theory based models; or
(2) phenomenological or parametric models.
Usually, phenomonological models are used because either no theory
exists or the interaction between several phenomena leads to com-
plexity in the theory whereas the overall behaviour may be easily
parameterised.
The real gas law:
pV = nzRT. (3.1.1)
Using a temperature and pressure dependent correction factor, z, is a
phenomenological modification of the ideal gas law which is a model
based on the classical kinetic theory of gases.
109
An alternative classification of models is:
(1) physical models; or

(2) mathematical models.
Usually, the results of physical models are used in order to make

a parametric or phenomenological description of the process being
studied. Newton’s laws of motion and Planck’s law of radiation are
two examples of laws derived by these methods.
The procedure of experimental observation and parametric anal-
ysis has been the basis for advance in many branches of the physical
sciences since the time of alchemists and indeed, application of
models can be said to have been the major factor in transforming
disciplines from arts into sciences.
Although, in this course, interest is focused on mathematical
models of hydrocarbon reservoirs, it is important to consider the
physical models in order to understand their role in reservoir
simulation.
3.1.2. Physical Models

Physical models may be divided into three categories:
(1) Analogue models;

(2) Comprehensive models; and
(3) Elemental models.
3.1.2.1. Analogue models

The basis of analogue models is the study of a phenomenon which
has similar behaviour to the one of interest, in order to investigate
the dependence on parameters which have analogues in the system
of interest.
Examples of such models are:
(1) heat flow to simulate single phase fluid flow; and

(2) electrostatic experiments to simulate flux lines between wells.
These models are little used in reservoir simulation.

Numerical Simulation 111
3.1.2.2. Comprehensive models

Comprehensive models are those in which all the important facets of
the reservoir production are included in a physical model. Because
of the size of the problem most of these are scale models.
In scaled models, reservoir dimensions, rock properties, fluid
properties and velocities must be scaled so that proportionally, the
forces in the model conform to those in the reservoir. In practice
this is difficult, if not impossible, to achieve as, for example, scaling
down the permeability would usually have the effect of scaling up
the capillary pressure hence altering the ratio between capillary,
gravity and viscous forces. Such alteration would render the results
of the model inappropriate for direct application to field scale
problems.
An example of a comprehensive model is the vertical displace-
ment of oil by gas in a long core at reservoir conditions. Here,
the reason for performing the experiment is a lack of ability to
adequately represent the interaction of the displacement and mass
transfer effects.
Although these models are useful and the results can be included
in reservoir simulation models, they have very little predictive power
as the basic mechanisms are not parameterised.
3.1.2.3. Elemental models

Elemental models are constructed as a microcosm of a certain aspect
of the reservoir behaviour and as such the experiments performed on
reservoir core plugs and fluid samples are examples of this genre. By
their nature these experiments concentrate on a restricted range of
phenomena actually occurring in the reservoir and attempt to isolate
these from the possible interactions between the studied phenomena
and other effects. Because of this requirement, these experiments
must be carefully designed to specifically exclude these extraneous
effects.
An example of such a model is the behaviour of oil and gas flowing
in a porous medium giving the experimentally observed relative
permeability. As the purpose of the experiment is to quantify the
multi-phase flow characteristics, the experiment is designed to elim-

inate any effects of mass transfer between phases and gravitational
segregation of fluids.
Depending on the fluid characteristics and production mechanism
of the reservoir, this “elemental model” might or might not be
directly applicable to the flow of fluids in the reservoir.
3.1.3. Mathematical Models

A mathematical model of a system is simply an equation which
relates the behaviour of the system, expressed in terms of observable
variables, to some parameters which describe the system. These
equations are frequently described as physical “laws”.
Apart from very few fundamental laws, all of these are gener-
alisations with some restricted area of validity. Examples of such
generalisations are Newton’s laws of motion, the ideal gas law and
Darcy’s law.
As the physical complexity of any system increases, the restric-
tions imposed, in order to produce an analytically or easily soluble
mathematical model, increase. The danger in using any mathematical
model is that if it is used outside the range imposed by the
assumptions inherent in its formulation, the results may appear
unphysical or, worse, they may not appear unrealistic but might
give a completely false picture of the important mechanisms and
the overall behaviour.
Examples of mathematical models applied to petroleum reser-
voirs are decline curve analysis and material balance. These tech-
niques are very useful in characterising well response in an unchang-
ing production scheme and overall reservoir and aquifer behaviour
but, because of the simplifying assumptions, they are of less use for
detailed reservoir description purposes or for formulating remedial
action for production problems.
3.1.4. Numerical Models

A more detailed mathematical model can be constructed by subdi-
viding the reservoir into small volume elements and applying the laws
of mass conservation and fluid flow to each element. By letting the

elements tend to zero volume, the equations for movement of fluids
in a porous medium can be constructed. For the formulation of these
equations see the section on Equations and Terminology.
The resulting equations are nonlinear partial differential equa-
tions which are almost always too complicated to solve analytically
even for a simple description of the hydrocarbon fluid properties.
Further approximations are made in order to solve the equations
at discrete points in space and time and it is this discretisation
which leads to the requirement to solve large linear matrix systems.
The availability of high speed digital computers together with
mathematical research has rendered this problem feasible even for a
very large matrix of space points (corresponding to a high resolution
in the reservoir description). The formulation and solution of these
equations is of importance to the reservoir simulation engineer using
the model so that the effects associated with the formulation and
numerical solution are understood.
Examples of such numerical effects are saturation oscillations due
to the explicit dating of mobility terms or the non-convergence of
fully implicit models due to discontinuities in the input data.
3.1.5. Models as Comparative Tools

Reservoir Engineers can use numerical reservoir simulation as a tool
to improve their insight into the behaviour of the fluids in the reser-
voir and therefore lead to better reservoir management. However,
any reservoir model does not simulate the processes of hydrocarbon
production from a reservoir but represents an approximate solution
to a mathematical simplification of the very complex reality.
Much of the data, particularly regarding the description of the
reservoir away from well control points, is the result of geophysical,
geological and engineering interpretation which can lead to the
basic model, as initially conceived, giving rise to significant errors
in simulated well or field performance. Thus, at an early stage of
development, models may be used to quantify the sensitivity of
forecasts to some poorly determined parameter by making several
predictions each with a different value and, at a later stage when
production data becomes available, the model may be adjusted in

order to better represent the observed trends.
This refined reservoir description may then be used for optimising
hydrocarbon production.
In either case, the model is used in a comparative way, either
comparing model with model to investigate sensitivity, or comparing
model with field results in order to improve the reservoir description
so that meaningful predictions may be made.
3.2. Equations and Terminology

The individual basic “laws” governing the flow of fluids in a porous
medium give rise to simple equations. When these equations are
combined and coupled with a simple description of the hydrocarbon
phase behaviour together with the effects of immiscible displacement,
the result is three coupled nonlinear second order partial differential
equations for which, in general, no analytical solutions exist.
In order to reach an approximate solution, the technique of finite
differences is (almost exclusively) used and this gives rise to a very
large set of linearised equations which must be solved.
The formulation of the finite difference scheme and the solution of
the linearised matrix equation using digital computers may produce
numerical effects which have important consequences for the practical
engineering results of the model.
Awareness of the origin of effects such as numerical oscillation,
dispersion of phases or components and the relationship between
input data and model performance can significantly change the
approach to and the results from a simulation study.
For this reason, an understanding of the techniques used and
their potential for affecting the simulation is of some importance to
the engineer when setting up and running a reservoir model.
3.2.1. Mass Conservation

The law of mass conservation can be applied to a small volume
element with fluid flowing through it along the direction of the x-
axis. This is illustrated in Fig. 3.2.1.
MASS BALANCE
x x + dx
x u
x x x y z t x ux y z t
x y z( t t t )
( u) ( )
x t
Figure 3.2.1. Mass balance.
Applying mass balance at the input and output faces gives
mass in − mass out = accumulation
ρx ux δyδzδt − ρx+δx ux+δx δyδzδt

= φδxδyδz(ρt+δt − ρt ). (3.2.1)
Cancelling terms and taking limits ∂x → 0 and ∂t → 0 gives

∂(ρu) ∂(ρφ)
= . (3.2.2)
∂x ∂t
This equation expresses the conservation of mass for a fluid moving
in a 1D system.
3.2.2. Darcy’s Law

In the middle of the last century, Darcy performed a set of exper-
iments in order to calculate the size of sand filtration beds for the
water supply of Dijon. A schematic diagram of his apparatus is shown
in Fig. 3.2.2.
He discovered that the rate of flow of water through his equip-
ment was linearly related to the difference in manometer height
k
u ( p gz )
l
Figure 3.2.2. Darcy’s law.
between the inlet and outlet.
uw ∝ (h1 − h2 ). (3.2.3)
This result was generalised for other experimental configurations and

fluids, yielding Darcy’s law
k ∂
u= (p + ρgz). (3.2.4)
µ ∂l
It is worth noting that this equation violates Newton’s Laws of

Motion, as a particle of fluid in a constant potential gradient moves
with a constant velocity.
This is similar to the problem of the terminal velocity of a
block moving down an inclined plane where gravitational forces are
dynamically balanced by frictional forces (see Fig. 3.2.3).
Darcy’s law implicitly contains a dynamic balance between the
measured potential acceleration force and complex retarding forces.
These retarding forces actually depend on the nature of the fluid flow
and because of this deviations from this law occur at high flow rates
where turbulence effects become important.
α v3
mgSinθ
Figure 3.2.3. Dynamic equilibrium.
3.2.3. Diffusivity Equation

Generalising the mass conservation and transport equations to 3D
gives
∂
∇(ρu) = (φρ) (3.2.5)
∂t
and
k
u= ∇(p + ρgz). (3.2.6)
µ
Removing the gravitational term for the moment and combining
these equations gives
ρk ∂
∇ ∇p = (φρ). (3.2.7)
µ ∂t
Assuming constant porosity and small pressure gradients in a low
compressibility fluid, that is
∂ρ ∂ρ ∂p ∂p
= = ρc . (3.2.8)
∂t ∂p ∂t ∂t
Then
φµc ∂p
∇2 p = . (3.2.9)
k ∂t
This linearisation gives the diffusivity equation with the diffusivity
constant
k
cdiff = . (3.2.10)
φµc
Note that using cylindrical coordinates to represent flow towards a

well gives

1 ∂ ∂p φµc ∂p
r = . (3.2.11)
r ∂r ∂r k ∂t
Which is the radial form of the diffusivity equation, the solution

of which is used for analysis of well test data and some analytical
aquifers.
In fact, for a black oil system (oil and gas being the two
hydrocarbon components) we have the following six equations:
kko ∂
∇ ∇(po − ρo gz) − qo = (ρo φSo ), (3.2.12)
µo Bo ∂t
kkg kko
∇ ∇(pg − ρg gz ) + ∇ Rs (p)∇(po − ρo gz) − qg
µg Bg µo Bo
∂
= {(Rs ρo So + ρg Sg )φ}, (3.2.13)
∂t
kkw ∂
∇ ∇(pw − ρw gz) − qw = (φρw Sw ), (3.2.14)
µw Bw ∂t
So + Sw + Sg = 1.0, (3.2.15)
po = pcow (Sw ) + pw , (3.2.16)
po = −pcog (Sg ) + pg . (3.2.17)
These six equations and six unknowns represent the model for fluid
flow in the reservoir but, due to coupling (Rs depends on p, pw
depends on po as well as pcow (Sw) and the nonlinearity, (ko depends
on So , Bo depends on po ) in general no analytical solution is possible.
However an approximate solution can be obtained for these
equations at preselected space points and for discrete time values
by using the method of finite differences.
3.2.4. Finite Difference

The method of finite differences makes use of an approximate form
for the derivative in order to evaluate the difference equation for each
selected space point. For simplicity, consider a function f (u), shown
in Fig. 3.2.4, on a uniform grid u0, u1, . . . , un − 1, un where
ui+1 − ui = δu . (3.2.18)
Using Taylor’s theorem we can write
∂u2 ∂u3

f (ui+1 ) = f (ui ) + ∂ufi + f + f + ··· . (3.2.19)
2! i 3! i
∂u2 ∂u3
f (ui−1 ) = f (ui ) − ∂ufi + f − f + ··· . (3.2.20)
2! i 3! i
Using this to approximate derivatives we obtain
∂f f f (ui+1 ) − f (ui )
= + O(δu), (3.2.21)
∂u δu
∂f b f (ui ) − f (ui−1 )
= + O(δu), (3.2.22)
∂u δu
∂2f f (ui−1 ) + f (ui+1 ) − 2f (ui )
= + O((δu)2 ), (3.2.23)
∂u2 δu2
i+1
i
f (u)
i 1
Figure 3.2.4. Finite difference.

where the superscripts b and f refer to backwards and forwards as

the directions for computing the gradients.
If we apply these approximations to a simplified oil equation in
one space dimension, then we obtain

1 kko kko
(pi+1 − pi ) − (pi − pi−1 )
δx2 µo Bo i+1/2 µo Bo i−1/2

1 φSo n+1 φSo n
−q o = − , (3.2.24)
δt Bo i Bo i
where for each space point we have a similar difference equation for
each of the three fluid components.
Multiplying through by the volume of the element converts the
specific rate term to mass per unit time and gives:

δyδz kko kko
(pi+1 − pi ) − (pi − pi−1 )
δx µo Bo
i+1/2 µo Bo i−1/2

V φSo n+1 φSo n
−qo = − . (3.2.25)
δt Bo i Bo i
The superscript n refers to the start of the time step and n + 1 to

the end of the time step.
Thus because of the errors inherent in the approximation of the
derivatives, for each space point and in each component, errors are
introduced into both the space and time parts of the equations.
These errors are called truncation errors.
The space truncation errors being of order δx2 are usually small
provided that the grid of space points is selected to be reasonably
regular. For adjacent grid cells with large differences in spatial extent,
space truncation errors can become important.
More often, time truncation errors are more important as the
derivative approximation is correct only to order δt and the error
grows with the time step size.
3.2.5. Implicit and Explicit Formulation
Figure 3.2.5. Flow between two cells.
EXPLICIT FORMULATION
Ak ∆p 1
δρ = δt ,
µo Bo ∆x A∆xφ
∆x2 φµo c
δt ≺ . (3.2.26)
2 k
We have written for the oil equation

δyδz kko kko
(pi+1 − pi ) − (pi − pi−1 )
δx µo Bo
i+1/2 µ o Bo i−1/2

V φSo n+1 φSo n
−qo = − . (3.2.27)
δt Bo i Bo i
This difference equation is formulated to obey Darcy’s law and to

conserve mass. The computed flux between two space points will
depend on the dating of the mobility term and the pressure difference
term.
If we consider two cells completely filled with oil with an initial
pressure difference ∆p (shown schematically in Fig. 3.2.5), then from
Darcy’s law we can compute the instantaneous flow between cells
Ak ∆p
N= δt (3.2.28)
µo Bo ∆x
and the resulting change in density is
Ak ∆p 1
δp = δt . (3.2.29)
µo Bo ∆x A∆xφ
For a total compressibility of c, then the change in pressure in the

cell as a result of fluid flow is:
k ∆p
δp = δt. (3.2.30)
φµo c ∆x2
For stability, δp must be less than ∆p/2

or
∆x2 φµo c
δt < . (3.2.31)
2 k
Using:
∆x = 104 cm
µo = 1cp
c = 10−4 atm−1
k = 1darcy
Φ = 0.2
then δt < 103 s or approximately 15 min.

If this stable time step is exceeded, then oscillations (possibly
amplified) will occur.
In order to relax the stability criterion to provide useful time step
size, the pressure must be treated implicitly. The difference between
the explicit and implicit flux calculation is illustrated in Fig. 3.2.6.
Unstable
Explicit
q12
Stable
Explicit
Stable Implicit
Time 1 Day
Figure 3.2.6. Implicit and explicit.
Writing the pressure difference at the new time level, we have

∂y∂z kko
(pi+1 − pi )n+1
δx µo Bo i+1/2

kko
− (pi − pi−1 )n+1
− qo
µo Bo i−1/2

V φSo n+1 φSo n
= − , (3.2.32)
δt Bo i Bo i
Linearising the changes in Φ and 1/Bo with respect to pressure and

writing

k
bi = −δtδyδz (pn − pni )
µo Bo i+1/2 i+1

k
− (p − pi−1 ) ,
n n
(3.2.33)
µo Bo i−1/2 i

k
ai− = δtδyδz, (3.2.34)
µo Bo i−1/2

k
ai+ = δtδyδz, (3.2.35)
µo Bo i+1/2
 
1
1 ∂φ ∂ Bo
ai = a− + a+ +  −φ  V δx. (3.2.36)
Bo ∂p ∂p
Then we can write
bi = ai− ∆pi−1 + ai ∆pi + ai+ ∆pi+1 , (3.2.37)
where the change in pressure over the time step is
∆pi = pn+1
i − pni . (3.2.38)
For a 1D problem with m space points we have m such equations

    
a1 a1+ · · · · ∆p1 b1
   b 
a2− a2 a2+ · · ·   ∆p2   2
    
 · · ·     b3 
 a3− a3 a3+   ∆p3   
  =  (3.2.39)
 · · · · · ·     · 
  ·   
 · · · · · · 
   · 
  ·   
· · · · am− am ∆pm bm
or in a more compact form
A∆p = b, (3.2.40)
where A is the coefficient matrix and ∆p and b are vectors of

length m.
The value of the pressure change for any individual block can no
longer be computed in isolation but its value is implicitly included
within the solution of the complete set of equations.
This leaves open the question of the dating of the mobility terms.
Depending upon the simulation problem, this might be stable using
schemes explicit in saturation.
Fully implicit schemes evaluate all coefficients on the left-hand

side of the equations at the new time level n + 1.
IMPES schemes (implicit pressure, explicit saturation) evaluate
saturation dependent terms at the start of the time step. There are
many schemes which lie between these in terms of the degree of
implicitness of the formulation.
Explicit schemes are quick to evaluate per time step and require
little computer storage space but have stability constraints which
may lead to oscillations or divergence of the solution.
Implicit schemes are slower per time step but are uncondition-
ally stable and for this reason have become the norm for black oil
models.
3.2.6. Dispersion and Weighting

The mobility term in the difference equation refers to the mobility of
the fluids at the interface between cells. In order to compute this we
would need the saturation and pressure at the cell interface which
by definition is not one of the calculation nodes. The usual approach
is to assign the mobility at the face to be that computed at the cell
centre of the upstream (higher potential) block.
This leads to fluid within the upstream finite difference cell being
dispersed and can result in poor simulation of sharp saturation (or
component) fronts (see Fig. 3.2.7). The severity of the numerical
dispersion which is produced depends on the size of the finite
difference mesh chosen.
Theory
Theory
Wcut
Sw
Simulation Simulation
x Time
Figure 3.2.7. Dispersion.

3.2.7. Nonlinearity and Outer Iterations

In order to set up the discretised equation for fluid flow
A∆p = b (3.2.41)
assumptions were made regarding the linear dependence of porosity

and the inverse of the volume factor with respect to pressure.
A consequence of this discretisation is that the value of B n+1
o is
required.
The technique used to compute this is a Newton–Raphson
approach where

1 1 ∂1/B o k k+1
= + ∂p . (3.2.42)
B k+1
o B(pk ) ∂p
The superscript k refers to the iteration level and δpk+1 is the

pressure change over the iteration.
The estimate of the value at the end of the iteration is expressed
as a linear function of the change in pressure over that iteration.
The first iteration uses the start of time step values for the initial
start point and gradient.
At the end of the first iteration, 1/Bo and its derivative are
re-evaluated at the new pressure value. This iteration procedure
k
1 1 1 / Bo
pk 1
Bok 1
Bo ( p k ) p
Figure 3.2.8. Newton iteration.

stops when
• all the values of δp are smaller than some tolerance, and
• the material balance is within a tolerance, or
• the maximum number of iterations has been reached.
The iterative procedure is illustrated in Fig. 3.2.8.
If, for some cells, the value of the formation volume factor at the
new time level, B n+1
o , is not convergent, then the densities used in
the difference equation will be incompatible and the solution will not
conserve mass. This will result in a material balance error.
3.2.8. Linear Solvers

Having written the equation
A∆p = b, (3.2.43)
the solution must be obtained for the vector of changes to the
pressure. For reservoir simulation, special purpose linear solvers have
been formulated in order to maximise the efficiency of this task.
This effort has been necessary because of the size of the problem
which for a fully implicit black oil model with one thousand cells
would contain nine million matrix elements if the A matrix was
treated as a general square matrix.
The solution algorithms developed take account of not only
the sparseness of the matrix but also of the regular nature of the
occurrence of matrix elements.
The actual location of the matrix elements depends on the
indexing scheme adopted for the finite difference space points and
this in turn can affect the number of arithmetic operations needed
to invert the matrix.
Some indexing schemes are shown in Figs. 3.2.9 and 3.2.10.
If the A matrix can be written as
A = LU, (3.2.44)
where L is lower triangular and U is upper triangular, then the
problem may be written
LU ∆p = b (3.2.45)
1 2 3 4 5
6 7 8 9 10
11 12 13 14 15
Figure 3.2.9. Matrix indexing scheme 1.
1 4 7 10 13
2 5 8 11 14
3 6 9 12 15
Figure 3.2.10. Matrix indexing scheme 2.

or
U ∆p = L−1 b (3.2.46)
and hence the back substitution
∆p = U −1 L−1 b. (3.2.47)
This corresponds to performing a forward elimination and a back

substitution giving the vector of pressure changes precisely.
The process of elimination involves manipulation of matrix rows
and requires a number of arithmetic operations dependent on the
difference in index between the outermost matrix elements in each
row. This is referred to as the “matrix band width” and is related to
the ordering scheme used and the number of cells in each dimension.
For a model with nx , ny and nz cells in the x, y and z direction
respectively, ordering first in the x direction gives a band width w
w = 2ny nz + 1. (3.2.48)
The classical approach of forward elimination and back substitution,

also known as Gaussian elimination, is called a direct solution. The
number of arithmetic operations involved in such a solution depends
on the band width and is approximated as
N ∝ nx n3y n3z . (3.2.49)
A different approach is to attempt an approximate LU decomposition

by iterative means. The number of arithmetic operations depends
on the formulation of the iterative technique but can be approxi-
mated as
N ∝ niter nx ny nz , (3.2.50)
where niter is the number of iterations required to converge to the

linear solution.
Although it is beyond the scope of this course to treat the many
iterative schemes, the fundamentals may be illustrated by considering
one such scheme.
For a 2D problem, we can write

A∆p = b (3.2.51)
as
     
T1 H1 · · x1 B1
     
G2 T2 H2 ·  x2  B2 
  =   =  , (3.2.52)
 · H3     
 G3 T3  x3  B3 
· · G4 T4 x4 B4
where Ti is a tri-diagonal matrix of dimension , where is the length

of the model. Hi and Gi are diagonal matrices of dimension and
xi and Bi are vectors of length of the change in pressure and b
respectively.
One method of resolving this set of equations is to make a direct
solution of the tri-diagonal submatrices as follows
xm+1
1 = T1−1 (B1 − H1 xm
2 ), (3.2.53)
xm+1
2 = T2−1 (B2 − G2 xm+1
1 − H2 xm
3 ), (3.2.54)
xm+1
3 = T3−1 (B3 − G3 xm+1
2 − H3 xm
4 ), (3.2.55)
xm+1
4 = T4−1 (B4 − G4 xm+1
3 ). (3.2.56)
Here m is the index of the iteration level for the linear solver.
This iterative procedure is repeated until all the elements of the
vector are smaller than some tolerance.
(xm+1 − xm ). (3.2.57)
The method illustrated here uses the latest up dated values of
the solutions of x in order to modify the right-hand side and is known
as a line Gauss–Seidel procedure.
This iterative procedure is at the level of the solution of the
linearised equations and consequently the iterations, index m, are
referred to as “inner iterations” compared to the “outer iterations”,
index k, required to resolve the nonlinearities.
The overall iterative scheme is shown in Fig. 3.2.11 where the
relationship between inner and outer iterations can be seen.
Compute matrix
elements
Set up linearised
matrix equation
Outer
Perform 1 Gauss-Seidel Iterations
pass
Inner
No Iteration
Check for convergence
of linear solver
Yes
No
Check for convergence
of
of non-linear
nonlinear effects
effects
Yes
Output solution
Figure 3.2.11. Overall iteration scheme.
3.3. Buckley–Leverett Displacement

If we consider the flow of an incompressible fluid (water say)
displacing a different, immiscible fluid from a 1D porous system at
constant fluid rate, then applying conservation of mass gives

∂qw ∂Sw
= −Aφ . (3.3.1)
∂x t ∂x x
We can write the differential for the water saturation as

∂Sw ∂Sw
dSw = dx + dt. (3.3.2)
∂x t ∂x x
If we consider the movement of a particular water saturation contour

through the system then for that contour dSw = 0. Then

∂Sw ∂Sw dx
=− . (3.3.3)
∂t x ∂x dt Sw
Defining the fractional flow of water as

qw
fw = , (3.3.4)
qtotal
which for water–oil displacement is
qw
fw = , (3.3.5)
qw + qo
then, because the total flow rate is constant,
∂qw ∂fw
= qtotal (3.3.6)
∂x ∂x
and
∂fw ∂fw ∂Sw
= . (3.3.7)
∂x ∂Sx ∂x
Substituting Eqs. (3.3.3) and (3.3.7) into Eq. (3.3.1), we obtain

∂fw ∂Sw ∂Sw dx
qtotal = Aφ . (3.3.8)
∂Sw ∂x ∂x dt Sw
Cancelling and rearranging gives

dx qtotal dfw
νSw = = . (3.3.9)
dt Sw Aφ dSw Sw
That is, the velocity of the contour is proportional to the gradient of

the fraction flow at that water saturation.
The water fractional flow for a typical North Sea oil–water
system is shown in Figs. 3.3.1 and 3.3.2. This function results in
a mathematical inconsistency because the velocity of the high water
saturation is greater than the velocity of the leading lower saturation
contours.
Fractional flow of water
Fw
Sw Sw shock
Figure 3.3.1. Fractional flow.
Sw shock
Sw
Figure 3.3.2. Buckly Leverett displacement.
This results in the formation of a shock front or piston of water

moving through the 1D system and leads to an efficient displacement
of the oil.
This analytical treatment can be used to compare with the results
of numerical simulation in order to investigate the effects of numerical
dispersion in simple systems.
It should be recognised that although in displacement of water
by oil in oil fields there are dispersive contributions from capillary
pressure, vertical and lateral inhomogeneities of porosity and perme-
ability as well as the effect of gravity, these dispersive terms do not
behave in the same manner as the numerical dispersion.
3.4. Reservoir Models

Although all reservoirs are different and so the models built to study
them are different, the components of the basic input data and the
category of model are similar between all reservoir simulators.
3.4.1. Model Components

All reservoir models have components of data which correspond to
the major activities performed during reservoir simulation.
3.4.1.1. Reservoir description

The description of the reservoir requires
• the reservoir geometry,

• the rock properties,
• the multi-phase fluid flow properties,
• the fluid-phase behaviour properties,
• the aquifer properties.
3.4.1.2. Initialisation
In order to set the initial phase saturations and pressures, the model
requires
• the position of fluid contacts,

• the capillary pressure function,
• any spatial variation of fluid properties,
• a reference pressure at a specific depth.
3.4.1.3. Model control

Although frequently set by default, models have a large range of
control parameters which determine
• the finite difference formulation,

• the linear solution method,
• the number of nonlinear iterations,
• the required accuracy of the solution.
3.4.1.4. Production data

This consists of the time dependent data specifying
• well location and type,
• perforation interval and parameters,
• well target rates,
• well or group constraints,
• actions resulting from some violation.
3.4.1.5. Output
The user can request different types of reports or data to be output
• time step summary,
• well summary,
• region summary,
• cell by cell values of variables,
• graphics files,
• pseudo files,
• restart files.
The detailed form of the input data depends on the type and
input format specification of the model which is being used. As the
complexity of the physical processes treated by the model increases
so does the quantity of data require to define the processes. Although
most simulators provide some sort of data screening, they are by no
means infallible and the requirement for consistency in the input data
is the responsibility of the engineer setting up the model.
3.4.2. Model Types

The approach to reservoir modelling is one by which specific
behaviour of the field (or part of the field) may be studied in
some detail and simplified in some way so that the same behaviour
can be accurately preserved in a model which would not, a priori,
have sufficient resolution to simulate the effect. However, this lower
resolution model can include many wells and field constraints which
it would not have been feasible to be included in the initial high
resolution model.
The principal types of model used are described below.
3.4.2.1. Cross-sectional models

One of the major controlling influences on the production from oil
and gas reservoirs is the distribution of the permeability through the
reservoir section and the degree of vertical communication between
the reservoir layers.
The presence of a thin but continuous layer of very high or low
permeability may have an effect on the production characteristics of
the field (or sector) which must be faithfully reproduced in a model
where the layer thickness could be many times the magnitude of the
particular zone.
A cross-sectional model permits the production characteristics
to be investigated and for the engineer to represent the resulting
reservoir behaviour in a form suitable for inclusion in the coarse
model.
In certain fields, a representative cross-section may be used to
investigate future production and the results scaled in some way to
give the expected field behaviour.
3.4.2.2. Sector models

The use of a representative cross-section whose results are scaled to
give the behaviour of the field, is actually a special case of the sector
model where a representative microcosm is studied in some detail
and the field profile is computed by integrating the results in some
way, including the effects of scheduling and constraints.
Models of symmetry elements of pattern floods and isolated fault
block models are examples of such sectors.
Sector models are usually run when the flux across the external
boundary of the model is zero or negligible. In certain cases it is
possible to define the fluxes in some way and perform a sector
model study with externally imposed flux values at each time step.
Such models have been used to study the effects of several oil fields
producing from a zone connected by a common aquifer.
3.4.2.3. Cylindrical single well models

Single well models are frequently used to improve the reservoir
description as a result of some effect in the pressure behaviour of a
well test. They are also used for investigating the dependence of the
produced fluids on rate and contact position either for optimisation
of the completion or for including the results in a coarse simulation
model.
3.4.2.4. Full field models

The individual high resolution models are used for detailed exami-
nation of a particular zone or effect, but the overall interaction of
production and injection in all other wells as well as the imposition
of group or field constraints is studied by using a full field model. The
high resolution model behaviour is reflected in the full field model
by the use of up-scaling to represent the fluid flow or production
characteristics of a well.
Despite the use of up-scaling to integrate the detailed reservoir
behaviour into a form for full field simulation, it is usual for the coarse
model to have many tens of thousands of active cells and models with
hundreds of thousands of grid blocks are quite normal.
Producing coherent input data and analysing the results of
models of this size is a major task.
3.5. Grid Systems

In Sec. 3.2.4, we wrote the finite difference form of the 1D, equally
spaced, Cartesian diffusivity equation for oil. Simplifying the nota-
tion, we can write this as

1 φSo
Txi+ (pi+1 − pi ) − Txi− (pi − pi−1 ) = ∆t , (3.5.1)
δt Bo
where T is the transmissibility term and ∆t is the time differ-

ence operator. In cylindrical coordinates, the linearised diffusivity
equation becomes

1 ∂ ∂p φµc ∂p
r = . (3.5.2)
r ∂r ∂r k ∂t
This may be discretised in the same way as the Cartesian equation,
giving

1 kko
r (pi+1 − pi )
ri δri µo Bo i+1/2 i+1/2

kko
− r (pi − pi−1 )
µo Bo i−1/2 i−1/2

V φSo n+1 φSo i
= − , (3.5.3)
δt Bo i Bo n
where r i−1/2 represents a point between point i and i − 1 where the

mobility term is computed. It is not clear that for radial coordinate
systems the arithmetic average location is a good choice. For steady
state inflow into a well it can be shown that the logarithmic mean
ri+1 − ri
ri+1/2 = (3.5.4)
in(ri+1 /ri )
is a better choice.
For some choice of transmissibility centre, we can write

1 φso
Tri+ (pi+1 − pi ) − Tri− (pi − pi−1 ) = ∆t (3.5.5)
δt Bo
which is identical in form to the Cartesian case, except that the
different geometrical configuration has been incorporated into the a
priori computation of the transmissibility equation.
This property is quite fundamental to most reservoir simulators
as (with minor modifications) more complex coordinate systems can
be dealt with in a standard simulator by a preliminary treatment of
the geometrical terms which affect the transmissibility calculation.
Note that whilst a constantly spaced 1D Cartesian system and
a logarithmically spaced 1D radial system are completely analogous,
a 2D areal Cartesian model and a 2D areal radial model are not.
This is due to the cyclic nature of the boundary conditions for

cylindrical models, discretised in the azimuthal coordinate, because
of the connection between the first and the last sectors.
3.5.1. Cartesian Systems

Cartesian models (see Fig. 3.5.1), built on rectangular grid cells, are
one of the most common type of models. This is because they are
easy to set up and the results are easily interpreted. The potential
problems with these models are that the choice of gridding may not
be optimum and that grid orientation effects may be present.
Never the less, the ease of setting up the model, together with
the ease of relating the results to field behaviour are the main reasons
for these models being the first choice for many simulation projects.
The choice of the dimensions of the grid is a compromise between
accuracy and cost (or computer execution time). The size of a grid
for any particular model depends on:
(1) the definition required for the structural description,

(2) the location of fluid contacts,
(3) the proposed well density.
Figure 3.5.1. 3D Cartesion system.

Clearly the mesh size should also depend on the amount and quality
of data available. There is little point in constructing a detailed
model with many thousands of grid cells for a prospect with only
the discovery well and coarsely spaced 2D seismic data available.
3.5.2. Cylindrical Systems

A schematic cylindrical system is displayed in Fig. 3.5.2. As
demonstrated above, the mathematical formulation (apart from
cyclic boundary conditions) is independent of the coordinate system
chosen. Historically this was not the case, as the numerically more
sensitive cylindrical coordinate models were frequently designed to
have a higher degree of implicitness which could not feasibly be run
on a large scale 3D model.
These single well models (frequently called coning models,
because the study of coning phenomena was one of their primary
uses) have largely been incorporated into the standard commercially
available systems as an option for the coordinate system choice.
This is possible because the current generation of black oil reservoir
Figure 3.5.2. 3D Radial cylindrical system.

simulators have available the option for a fully implicit formulation

of the difference equations.
3.5.3. Stream Line Grids

In the same way that mapping the radial system on to the Cartesian
system with a one-to-one correspondence between grid cells, permits
the simulator to operate independently of the coordinate system,
once the pore volume and transmissibility values have been correctly
calculated, any system which preserves the cell indexing and inter
connection scheme can be used provided that the appropriate
geometrical calculations have been made.
One such choice is the curved set of grids obtained from using
streamlines and iso-potentials for defining the grids. These grids
are however analytically calculable for a restricted set of well
configurations and fluid mobility ratios and maybe distorted by
reservoir inhomogeneity or changes in field operation. Because of
this, a simplified approach may be used approximating the flow lines
by straight lines. Such a system is shown in Fig. 3.5.3.
However, in certain circumstances they can provide a cost
effective solution to rather complex problems but great care must be
taken as the applicability of the results from the simulator depend
critically on the accuracy of the choice of stream lines.
Figure 3.5.3. Approximate stream line system.

3.5.4. Special Connections

The preservation of the mathematical mapping between the different
coordinate systems permits the simulator to operate because the
matrix coefficients can be computed, using predefined geometri-
cally dependent terms, and the location of the matrix elements is
unperturbed.
This is not quite true for the cylindrical system, where the
connection between the first and the last azimuthal sector requires
some additional matrix elements. These however appear in strictly
defined positions and can be accommodated in the efficient solution
of the linearised equations.
Further additions to the matrix are necessary for the inclusion
of well constraints. If a well, which is perforated over several
non-communicating layers, is producing at it maximum permitted
rate then the contribution from each layer depends implicitly on the
others.
qtotal = ql1 + ql2 + · · · + qlk . (3.5.6)
This introduces matrix elements connecting all the perforated cells in

the well. The location of these additional matrix elements is clearly
less well defined than those produced by cyclic boundary conditions
and may also be time dependent as wells are perforated.
Geological pinch out also juxtaposes cells not adjacent in the
normal indexing schemes. These effects are shown schematically in
Fig. 3.5.4.
The treatment of flow across fluid transmitting geological faults
(see Fig. 3.5.5) can also introduce matrix elements which are
additional to the usual regular pattern. This is because a fault
having a throw equal to half of the thickness of a layer will reduce
the transmissibility in that layer but add transmissibility matrix
elements corresponding to the additional communications between
layers across the fault.
Although the occurrence of these elements is not (usually) time
dependent, they may be large in number for a model of a large faulted
field, and the disruption to the matrix pattern may be severe.
Cross flow
Pinch out
Figure 3.5.4. Well layer flows and pinch out.
Fault
Figure 3.5.5. Flow across faults.

This can have severe consequences for the speed of each inner
iteration and also on the convergence properties of the matrix.
3.5.5. Corner Point Representation

The extension of the mapping of the areal coordinate system to
a distorted grid, following the fault directions, as well as the
requirement to model non-vertical faults leads to grid cells which may
have a general six faced shape. It is no longer possible to describe the
cell geometry simply by its extension in 3D and the solution adopted
is to give the simulator the space coordinates of the eight points
corresponding to the corners. Note that the neighbouring grid block
is also specified using eight corner points and there is no requirement
that the points between the cells should be coincident.
Provided that the transmissibilities can be calculated between
neighbouring cells, then the distorted grid is treated by the simulator
in the same way as any other geometrical mapping.
3.5.6. Local Grid Refinement

The usual approach to the spatial discretisation is to choose a set of
grid cells defined by lines superimposed on the structure map. These
lines may be curved, if a non-cartesian system is being used, but, in
order to preserve the matrix structure, they are continuous.
The ability to deal with grid cells having more than the six
nearest neighbours was required for the treatment of flow across
geological faults and the same technique can be used to set up the
matrix corresponding to a locally refined grid. Examples of the use
of such grids are shown in Fig. 3.5.6.
Use of local grid refinement automatically adds special connec-
tions into the matrix. An example of this is shown in Fig. 3.5.7,
where the extra transmissibility values corresponding to both the
refined grid and its connection to the global grid may be seen.
Because the matrix structure may be seriously disturbed, the
time taken per linear iteration and the number of iterations required
for convergence may be increased substantially.
Oil zone
Transition Zone
Well
Aquifer
Figure 3.5.6. Local grid refinement.
1 2 3 4 5
6 7 8 9 10
15 16
11 12 13 14 17 18
Figure 3.5.7. Local grid refinement showing the effect on the matrix structure.
3.5.7. Unstructured Grids

As the complexity of the grid increases, then there is a corresponding
loss of simplicity in the matrix structure which when taken to
the limit means that the coefficient matrix becomes a generalised
sparse matrix with little if any evident structure. Such unstructured
grids may result from a gridding technique using perpendicular
bisectors (PEBI) of lines joining a set of control points. These control
points become the grid node and the geometry associated with the
node is determined by the set of PEBI lines surrounding each one.
Figure 3.5.8 shows an example of a PEBI grid. Notice that some
parts of the grid look almost Cartesian whilst some cells have five or
six nearest neighbours.
Figure 3.5.8. Example of unstructured pebi grid.
Gathering centre 1
Figure 3.5.9. Additional matrix elements due to surface constraints.
3.5.8. Surface Constraints

In the same way that complex gridding complicates the matrix struc-
ture, adding surface constraints can add many matrix coefficients not
normally present in the set of equations governing inter-block flow. A
simple example of this is shown in Fig. 3.5.9, where the constraint at
the manifold propagates back into the subsurface matrix connecting
many blocks with special connections.
Most present day simulators do not include the gathering centre
constraints in the matrix as the special connection overhead is too
great so the constraint is wound in through the Newton iteration
scheme. However, a new generation of simulators (Landmark’s nexus
and Schlumberger’s Intersect) built to handle unstructured grids can
include the surface system in the subsurface matrix.
3.5.9. Non-Orthogonality
In Sec. 3.5.3, we described the use of stream lines for setting up
the spatial discretisation. This system is useful only in so far as the
flow is along the stream lines and such a grid system would not
be chosen for a different application. This is because the choice of a
non-orthogonal system should lead to transmissibility terms between
cells not usually connected in the normal five point differencing
schemes. However, for the stream line choice, the flux in these
directions is zero so that neglecting these terms does not lead to
errors.
For an application with reasonably fixed flow patterns (for
instance the simulation of a symmetry element of a pattern develop-
ment) this procedure can be very useful but care must be taken that
field operating conditions do not seriously alter the stream lines.
Although distorted coordinate systems may give an improved
geometrical representation of the reservoir and its fault system, it
may also introduce a further level of divergence from the differential
equation due to neglecting some cross-term transmissibilities. In this
respect it is similar to an additional space truncation error.
The same is true of the case where local grid refinement is used
as the fluxes between the fine grid area and the coarse grid area are
subject to additional truncation error terms which depend on the
degree of refinement.
However, apart from trying different grid systems (which is not
frequently done), there is no a priori way of estimating the size of
the error introduced.
A final comment on orthogonality of reservoir models is that the
choice of axes, x along the dip, y along the strike and z vertical
is non-orthogonal, except for horizontal models, and that the error
introduced is usually small compared to all the other approximations
used and the accuracy of the data available.
3.6. Rock Properties

The porous medium containing the hydrocarbons is a physically and
chemically complex material whose properties and the dependence
of these properties on changes brought about by hydrocarbon pro-

duction determine, to a large extent, the production characteristics.
In black oil reservoir simulation, most of these properties are
simplified or ignored, the rock being represented as a chemically inert,
temperature independent medium characterised by its porosity, per-
meability, rock compressibility and saturation dependent functions
of relative permeability and capillary pressure.
These properties are estimated from measurements made in a
variety of ways, including core data, interpretation of well test data,
interpretation of log data and the use of correlations.
3.6.1. Core Data

The most direct measurements made are those made on the cores,
although even here, because the core has been brought to surface
and treated using some laboratory procedure, the results which are
obtained may be different from the in situ values.
Core analysis is broadly divided into the routine analysis,
performed rapidly soon after core retrieval, and the special core
analysis, which is much more time consuming, performed in order to
measure values of parameters used for petro-physical interpretation
and two-phase properties.
3.6.1.1. Routine core analysis

From the point of view of constructing a reservoir simulation model,
one of the most important aspects is to decide upon the layering
system to be used. This usually depends on the vertical distribution
of both vertical and horizontal permeability and good core coverage
is a very desirable starting point as the routine analysis provides
numerical values for the porosity and permeability on some constant
sampling interval.
3.6.1.1.1. Porosity
The porosity of interest to reservoir engineers is the value of
the connected porosity (which may be different from the absolute
porosity as measured by logs). This is rapidly measured by cleaning
and drying the core sample and comparing the weight of a helium
container with and without the core. This gives a value for the grain
volume and, knowing the value for the bulk volume, the connected
porosity can be calculated.
An additional estimate of porosity is often available from the
amount of fluids which were recovered from the core sample.
3.6.1.1.2. Permeability
The cleaned and dried core is inserted into a sleeve and a gas (usually
air) is passed through the core. The pressure difference is adjusted
to give a measurable flow rate from which the permeability can be
calculated. The dependence of rate on permeability for a low pressure
gas (density inversely proportional to pressure) becomes
kA
q= (p2 − p22 ), (3.6.1)
2pbase Lµ 1
where pbase is the base pressure and L is the length of the core plug.
The viscous flow of gases and liquids is not quite the same due to
the different scale of the boundary layers. Klinkenberg showed that
the gas permeability was related to the liquid permeability by

c
kg = kl 1.0 + , (3.6.2)
p
where c = constant and p = average pressure.

The liquid permeability can be obtained from the intercept of the
plot of gas permeability versus the inverse of the average pressure.
Usually the laboratory either quotes the necessary correction or
actually applies it before reporting. The effect is usually to reduce
the measured permeability by some 10%.
Darcy’s law is valid for viscous flow, but if the flow rate is too high
then an additional pressure drop due to turbulence results giving an
apparent decrease in permeability with increasing rate. This means
that permeability values in excess of 15,000 md are frequently not
reliable but such values are usually isolated and infrequent.
3.6.1.2. Geocellular up-scaling

Simulators expect not the overall effective permeability but the
average permeability of the net reservoir rock. Because of this, lateral
transmissibility calculation includes the net-to-gross ratio and if this
is already folded into the permeability up-scaling, the net-to-gross
will be multiplied twice instead of only once.
Several methods can be used to up-scale the grid block param-
eters from fine scale geocellular models to coarser finite difference
model grids. Great care needs to be exercised in scaling up the
permeability in a way that is consistent with the way that the sim-
ulator uses net-to-gross ratio and permeability in order to calculate
transmissibility values.
3.6.1.3. Special core analysis

The special core analysis experiments of interest here are those
dealing with the pressure behaviour of the rock and those dealing
with the effects of the presence of two immiscible phases.
3.6.1.3.1. Compressibility
In situ, rock is subjected to forces from the over burden and from the
fluid occupying the pore space. The difference between these is the
net confining pressure. As the pore pressure decreases, allowing rock
grains to expand, the net confining pressure increases, tending to
reduce the bulk volume of the rock. Both of these effects reduce the
porosity and this effect is included in the formulation of the finite
difference equations, usually as an elastic process with a constant
compressibility.
Experimentally, the core sample is subjected to triaxial loading
such that there is no deformation in the lateral direction and the
change in volume is recorded as a function of the net confining
pressure.
The values obtained should be used to compute the derivative of
porosity with respect to pressure at the initial reservoir net confining
pressure and corrected to give the uniaxial result by using the
Recompression
∆θ
Unloading
θ Burial
Grain pressure
Figure 3.6.1. Rock compressibility.
correction factor:

1 1+v
fcorr = ctriaxial , (3.6.3)
3 1−v
where v is Poisson’s ratio for the rock (approximately 0.3).
This gives a correction factor fcorr of approximately 0.6.
The effect of net confining pressure should be included in the
values of porosity and (sometimes) permeability which are input to
the model to correspond to the initial conditions.
If the reservoir to be simulated has a component of compaction
drive then the effective rock compressibility can be an order of mag-
nitude greater than the usual value (3 · 10−6 psi−1 to 7 · 10−6 psi−1 ).
The effect of compaction is not reversible so the standard elastic
treatment is not sufficient and a special option dealing with this
phenomenon should be used.
The effects of inelastic behaviour are demonstrated in Fig. 3.6.1.
3.6.1.3.2. Relative permeability

The presence of two immiscible fluids in the same pore space tends to
alter the mobility such that a modification of Darcy’s law is required.
This is written
Akkr (SΨ ) ∂
u(SΨ ) = (−p + ρΨ gz), (3.6.4)
µΨ ∂x
where Ψ is the index of the phase and kr (SΨ ) is the relative

permeability which is a function of the phase saturation, varying
between zero at the critical phase saturation to a value of one when
the core is completely occupied by the phase.
On further examination, it is clear that this expression is a
simplification as the behaviour of the fluid system when increasing
the wetting phase saturation (imbibition) is not the same as when
decreasing the saturation (drainage).
For practical purposes, the relative permeability of interest is
that dealing with the displacement of oil by either water or gas.
Typical oil–water and gas–oil relative permeability curves are shown
in Figs. 3.6.2 and 3.6.3, respectively.
For water as the displacing phase in a water-wet rock this is the
imbibition relative permeability having a dynamic range between the
irreducible water saturation and the residual oil saturation.
This function is measured in the laboratory by either steady
state or unsteady state measurements and can be performed at either
laboratory or reservoir conditions.
The standard special core analysis for a water–oil relative perme-
ability measurement is carried out using unsteady state displacement
Kro(Swc)
Kro Krw
Krw(1-Sorw)
Swc 1-Sorw
Figure 3.6.2. Water–oil relative permeability.

1.0
1.0
Krg(Swc)
Kro(Swc)
Krg
Kro
Kro(Sgc)
Krg(Sorg)
Sorg Sgc
Swc
Figure 3.6.3. Gas–liquid relative permeability.
at laboratory conditions and the fluids used are water (or simulated
formation water) and a standard mineral oil with a viscosity of
approximately 20 centipoise.
This choice of viscosities permits the initial conditions of connate
water saturation to be recreated in a short period of time and also
produces rapid water breakthrough so that the full range of the
relative permeability function can be measured.
This experiment is performed at high velocity in order to
minimise the effects of the capillary pressure discontinuity at the
entrance and exit faces of the core.
A more time consuming process is to measure the steady state
relative permeability by simultaneous injection of water and oil into
a core at rates corresponding to the velocity of fluids in the reservoir.
In order to remove the capillary end effects, long cores can be used.
This experiment may be carried out at laboratory conditions or at
reservoir temperature and pressure with live crude and simulated
formation water.
Because there is no longer a Buckley–Leverett displacement, the
relative permeability function can be measured for the equilibrium
core saturation which is set up as a result of the relative
permeabilities.
This experiment, apart from difficulties with emulsion formation,

requires time and a good deal of expense. However, there are
indications that the residual oil obtained from such experiments is
more representative for North Sea reservoirs than the values from
the standard unsteady state method.
3.6.1.3.3. Capillary pressure

Capillary pressure arises because of the interfacial tension (σ)
between two non-miscible phases. If one-phase wets idealised spher-
ical rock grains in the presence of the non-wetting phase, then the
pressure difference between the phases depends on the radii of the
lenticular wetting phase volume. This is illustrated in Fig. 3.6.4 for
idealised sand grains.
In the case of water and oil with water as the wetting phase, then
pc = po − pw , (3.6.5)
where pc is the capillary pressure given by

1 1
pc = σ + . (3.6.6)
r1 r2
This gives rise to the characteristic capillary pressure curve as a

function of wetting phase saturation. A typical capillary pressure
curve is shown in Fig. 3.6.5.
Sand Grain
Oil
r1
Water
r2
Figure 3.6.4. Origin of capillary pressure.

Figure 3.6.5. Capillary pressure.
The contact angle of the wetting phase actually depends on

whether the wetting phase is advancing or retreating (see Fig. 3.6.6).
This gives rise to a difference in the capillary pressure function for
imbibition and drainage, with the drainage curve having a higher
capillary pressure for any value of wetting phase saturation greater
than the irreducible saturation.
This phenomenon, called hysteresis, affects both the capillary
pressure, shown in Fig. 3.6.7, and the relative permeability, shown in
Fig. 3.6.8.
As will be shown later, the capillary pressure is one of the
determining factors in establishing an initial saturation distribution
in the reservoir.
For practical purposes the data which are required correspond to
the process of establishing the initial oil saturation which is usually
considered to be that of oil migrating into the reservoir; a drainage
process.
The capillary pressure then has a dynamic range between the
irreducible water saturation and one.
In the laboratory, capillary pressure is measured by applying
an external pressure and measuring the corresponding equilibrium
saturations of the wetting and non-wetting phases. The fluid
(a) (b)
Figure 3.6.6. Contact angle for different flow direction.
Pc
Drainage
Imbibition
0 Swc Sw 1.0
Figure 3.6.7. Hysteresis in capillary pressure.

Figure 3.6.8. Hysteresis in relative permeability.
combinations normally used are an air brine system, where air is the
non-wetting phase, or a mercury air system where air is the wetting
phase.
In the air brine experiment, air is forced into the core reducing
the initially high water saturation as the pressure increases. In the
mercury air experiment, mercury is forced into the core plug and the
saturation of mercury versus the mercury pressure is recorded.
The mercury injection experiment is frequently used as it is
rapid to perform but it has the disadvantage that the core sample is
permanently damaged and is of no further use.
3.6.2. Log Data

Electric wire line logs can measure quantities related to the in situ
porosity and saturation. These can be used to give porosity and water
saturation values and, in some cases, estimates of permeability.
An estimate of the porosity can be obtained from a comparison
of the density and neutron tool response. The density tool relies
on electromagnetic scattering of electrons in order to estimate the
density as the electron density is closely related to the density of
matter.
The neutron tool relies on the kinematic effect that the target
with the greatest effect is one with the same mass as the projectile.
This means that the neutron tool is sensitive to hydrogen atoms and
hence to hydrocarbons and water.
A combination of these tools then, indicating a reduction in
density and an increase in hydrogen content is indicative of a fluid
filled porous rock and the excursion of the curves is related to the
porosity and the fluid saturations. The response of the FDC–CNL
log for a porous interval is shown in Fig. 3.6.9.
The fluid saturations may be estimated from the resistivity
measurements because of the different electrical properties of drilling
mud filtrate, formation brine and hydrocarbons.
Measurement of the resistivity deep in the formation away from
the zone invaded by mud filtrate can give an estimate of the initial
saturations and the shallow resistivity can give an estimate of the
residual oil saturation. An example of the response to a water bearing
formation is shown in Fig. 3.6.9.
L PERM son VSHALE

0. 2 MD 2000 24 0 US/F 40 0 V/ V 1
RESDEP DEN SW PHIE

0. 2 OHMM 2000 1.95 G/ C3 2.95 1 V/V 0 1 V/ V 0
RESMED NEU PHIE BV W

0. 2 OHMM 2000 0.45 V/V -0 .15 1 V/V 0 1 V/ V 0
Figure 3.6.9. Neutron density and resistivity logs.

Wire line formation testing tools can, in principle, permit a

measurement of the formation permeability by analysing the pressure
build up in the sample chamber using theoretical solutions for
spherical flow. Practically, these data, although immensely useful
for the final pressure value, are little used for anything other than
qualitative information regarding the formation permeability.
A more widely used method of deriving permeability from logs
is to use a transform calibrated against core data from the field and
geological interval, usually of the form:
βSw Vclay φ
klog = αSw Vclay e , (3.6.7)
where αSw Vclay and βSw Vclay are the coefficients determined from the
log and core data cross-plots obtained from wells where correspond-
ing log and core data were available. An example of such a cross-plot
is shown in Fig. 3.6.10.
1000
100
Kair
10
1
10 14 18 22 25 30 34
Percent
Figure 3.6.10. Porosity versus permeability.

3.6.3. Test Data

The permeability measurement from cores may suffer from effects
due to the core handling and also represents a very small statistical
sample compared to the volume of the reservoir.
By contrast, a well test is sensitive to the in situ permeability
within the radius of investigation of the test and appears to overcome
both of these difficulties.
For a well producing at constant rate in an infinite, homogeneous
reservoir of constant thickness, a plot of the bottom hole flowing
pressure versus the logarithm of time would give a straight line whose
slope, m, is inversely proportional to the permeability
162.6qµBo
m= . (3.6.8)
kh
In practice, the build-up is used more frequently than the draw-
down data and both sets are frequently affected by afterflow,
lateral changes in reservoir characteristics, communication to non-
perforated zones and change of fluid type. A typical pressure build-up
plot is shown in Fig. 3.6.11.
Because of this, the extraction of coherent data from a well
test, as for the petro-physical data, is frequently a specialist task
involving considerable interpretation. An analysis performed without
consideration of possible variations in reservoir or fluid behaviour
may be seriously in error.
Pressure
Log (T+DT) / DT
Figure 3.6.11. Typical pressure build up.

3.7. Model Relative Permeability

In Sec. 3.6, we discussed the measurement of relative permeability.
The fact that such measurements are costly and time consuming
means that the samples for which data exist might not have the
required values of permeability and porosity for the model input.
Also, the experiments excluded other effects, such as gravity, and this
might or might not be representative of the actual displacement in
the reservoir. This section describes the manipulation of the relative
permeability and its use in developing pseudo relative permeability
tables and the corresponding layering schemes.
3.7.1. Data Manipulation

The experimental procedure used to establish relative permeability
relationships is specifically designed to exclude gravitational and
capillary pressure effects. Their normal use is then for models where
any such effects would be small within the grid cell which has the
relative permeability value assigned to it.
This implies that the cell size is small and in particular that
the layer thickness is small which is the case for detailed single well
models and for cross-sectional models.
Here the requirement to minimise the number of grid cells
to ensure that the problem is a tractable one is not usually a
constraint so that the user is free to subdivide the vertical section
into many layers so that cross-sections with hundreds of layers are
not uncommon.
Each layer has its permeability (horizontal and vertical), its net-
to-gross ratio and its porosity assigned. The model also requires a
relative permeability relationship (usually in the form of a table) but
it is unlikely that for every layer a representative core sample will
have been used for the experimental determination.
The technique recommended is to carefully review the relative
permeability data in order to select those experiments which appear
to have consistent results and then to plot the normalised relative
permeability curves.
This normalisation is carried out in two stages:

(1) the data are normalised over the dynamic range of the water
saturation,
(2) the end point values of the relative permeability are normalised.
For the case of water–oil data, the relative permeability is first
∗ where
transferred to a new normalised saturation variable Sw
∗ Sw − Swc
Sw = . (3.7.1)
1. − Swc − Sorw
The relative permeability values are then normalised according to
the end point values as follows:
∗)
kro (Sw
∗
kro (Sw )= (3.7.2)
kro (Swc )
and
krw (Sw∗)
∗
krw (Sw )= (3.7.3)
krw (1. − Sorw )
The curves resulting from this procedure for the oil relative perme-
ability are shown in Fig. 3.7.1.
Inspection of the resulting plots of normalised relative perme-
ability permits one or more representative or average curves to be
selected.
Different types of curve may be associated with say, different
values of net-to-gross ratio.
In order to use the curves as input to the model, the end point
values for the specific layer (or group of cells) must be specified.
These values can be estimated from inspection of cross-plots of the
relative permeability end points versus primary variables such as
permeability and porosity.
Usually a simple set of correlations (or values) suffices to give a
comprehensive set of relative permeability curves for the model e.g.
Swc = α − βφ, (3.7.4)

Sorw = γ, (3.7.5)
Figure 3.7.1. Normalised relative permeability.
kro (Swc ) = δ + log(k), (3.7.6)

krw (1. − Sorw ) = η + εφ, (3.7.7)
Examples of cross-plots for connate water saturation and residual oil

saturation are shown in Figs. 3.7.2 and 3.7.3, respectively.
These values can then be evaluated for each layer and
the corresponding table of denormalised relative permeabilities
constructed.
A similar exercise may be performed for the gas–liquid relative
permeability in order to obtain a corresponding set of tables. This
table then represents the effective permeability of gas and oil as
the oil saturation varies (keeping the water saturation fixed at the
connate value).
A consistency requirement is that the corresponding oil relative
permeability value (corresponding to Sw = Swc and Sg = 0) is the
same for both tables.
1000
100
Kair
10
1
0 20 40 60 80 100
Sw
Figure 3.7.2. Water saturation versus permeability.
1000
100
Kair
10
1
0 10 20 30 40 50 60
Sor
Figure 3.7.3. Residual saturation versus permeability.

3.7.2. Three-Phase Relative Permeability

Although the user enters only two tables each dealing with two-phase
flow, in the model some cells might have all three phases mobile.
At this point models make use of correlations to compute the
oil relative permeability, the most frequently used one being a
modification of Stone’s second correlation

krow
kro = kro (Swc ) + krw
krow (Swc )

krog
× + krg − krw − krg . (3.7.8)
kro (Swc )
An example of the behaviour of this correlation is shown in Fig. 3.7.4.

If all three phases have significant mobile saturations in a model,
care should be taken that this correlation is producing the type of oil
relative permeability expected. For most applications, gravitational
segregation of the phases results in an extrapolation into the three-
phase zone which is close to one axis so that the results are not
usually sensitive to the form of the correlation used. This is not the
Figure 3.7.4. Three-phase relative permeability.

case for thermal simulation where there is close proximity of all three
phases.
3.7.3. Vertical Equilibrium

If we consider a grid block with dimensions ∆x in the x direction
and ∆z in the z direction, shown schematically in Fig. 3.7.5, the
characteristic time for particle of fluid to traverse the block is:
∆x
τx = (3.7.9)
vx
and
∆z
τz = . (3.7.10)
vz
Flow across the block is determined by the viscous forces and flow
down the block by gravity forces, so that
∆x
τx ∞ dp
(3.7.11)
kx dx
and
∆z
τz ∝ . (3.7.12)
kz gδρ
If τx is much greater than τz then we can consider that within the
block, gravitational equilibrium is established effectively, instanta-
neously giving fluid segregation within the block.
Figure 3.7.5. Schematic of vertical equilibrium cells.

For an oil-water system, using oil field units, this corresponds to

∆x dp/dx
10 , (3.7.13)
∆z a
where a is the anisotropy ratio kv/kh.
If we assume this segregation occurs we can calculate the effective
permeability for water and oil moving out through the side of the
cell (not the top or bottom). Defining the transmissibility in the y
direction as
∆x∆z
Ty = k (3.7.14)
∆y
then for water we have
Tyw = Ty krw (Sw )

∆xzw
= kkrw (1 − Sorw ) (3.7.15)
∆y
and for oil
Tyo = Ty kro (Sw )

∆x(∆z − zw )
= kkro (Swc ). (3.7.16)
∆y
We can write the water saturation
∆zw
Sw = Swc + (1. − Swc − Sorw ) (3.7.17)
∆z
or
∆zw (Sw − Swc )
= (3.7.18)
∆z (1. − Swc − Sorw )
and substituting we obtain the effective water and oil relative
permeabilities as functions of water saturation:
Sw − Swc
krw (Sw ) = krw (1. − Sorw ), (3.7.19)
1. − Swc − Sorw
1. − Sw − Sorw
kro (Sw ) = kro (Swc ). (3.7.20)
1. − Swc − Sorw
Figure 3.7.6. Vertical equilibrium relative permeability.
These expressions result in the linear relative permeability relation-

ships, shown in Fig. 3.7.6, associated with the gravity segregation.
The measured shape of the relative permeability, corresponding
to the values at intermediate saturations is unimportant as, by
assumption, these saturations never occur.
Thus we see, that given some assumptions regarding the spatial
distribution of fluids in the grid block, we can compute a relative
permeability which is different from that which is measured.
Conversely, if we input the modified relative permeability func-
tions into the reservoir model, we are imposing the corresponding
spatial distribution of fluids inside the grid block.
3.7.4. Pseudo Relative Permeability

In the previous section, we saw how simple assumptions regarding
the dominance of one set of forces permitted an analytical calculation
of the effective relative permeability functions.
Other examples of such calculations are the pseudo relative
permeability functions which result from piston like displacement
in a set of layers. If we have n such layers which can be referenced in
the order in which they flood, then, when j layers have flooded, the
relative permeability values and average saturations are given by

j
i (1 − S i )
ki hi krw orw
i=1
k rw (S w ) = , (3.7.21)

j
ki hi
i=1

n
ki hi kro
i (S i )
wc
i=j+1
kro (S w ) = , (3.7.22)

n
ki hi
i=1
where

j
n
φi hi (1 − Sorw
i )+ φi hi Swc
i
i=1 i=j+1
Sw = . (3.7.23)

n
φi hi
i=1
If we assume that the flooding order is according to decreasing

values of permeability, then this gives Stile’s formula for pseudo
relative permeability. Conversely, if we assume flooding from the base
upwards, this gives the Deitz pseudo curves.
In the case where no such simple assumptions can justifiably be
made, then the results of a detailed cross-sectional model may be
used to attempt to group together some layers and represent their
joint behaviour by some pseudo function.
The existence of a coherent set of pseudo functions describing in
a simple way the behaviour of a group of layers is one of the main
bases for the vertical discretisation of the full field reservoir model,
the other prime consideration being an adequate description of the
perforated interval.
One simple method of calculating pseudo relative permeability
from a cross-sectional model is, for each group of cells, to calculate
the transmissibility weighted relative permeability. This is illustrated
in Fig. 3.7.7.
T x1 K ro1 T x 2 K ro 2
K ro
Tx1 Tx 2
Figure 3.7.7. Pseudo relative permeability.
For a vertical column of cells

n
Txi kro
i
i=1
kro (S sw ) = . (3.7.24)

n
Txi
i=1
This formulation can be derived by making the assumption that the

pressure gradient is the same in each layer, which is probably not
correct. Similar assumptions exist in the formulation of other pseudo
relative permeability functions. Because of this, all pseudo functions
are approximations and the test for whether they are adequate is to
resimulate the fine grid model with the coarse grid obtained from
combining cells. If the results of the model are the same to within
an acceptable (defined by the engineer) tolerance, then the pseudo
functions may be used in subsequent coarse models.
The pseudo functions, which reflect the spatial distribution of the
fluids, have been generated and tested with specific ratios between
the viscous and gravity forces. If they are to be used in a full field
model then the rate of production of reserves from the cross-section
should be set to correspond to that expected from the field.
Conversely, if major changes in operating policy mean that this
fractional off take changes significantly, it will be necessary to check
that the pseudo functions are reasonable for the new velocities. If
not, new functions should be generated and checked.
Figure 3.7.8. Rate dependence of pseudo curves.
The effect of modification of operating conditions on the location

of fluids is shown schematically in Fig. 3.7.8.
The definition of the pseudo relative permeability used above
was for combining cells in a vertical column but cells may also be
combined in the longitudinal sense (see Fig. 3.7.9).
The calculation of the average saturation is straight forward but
there is some choice for the selection of the relative permeability.
On the basis that transmissibility is truly related to the face
between grid cells, the logical choice is the relative permeability of
the cell closest to the outgoing face.
This gives rise to relative permeability functions shown in
Fig. 3.7.10 which retard the advance of the displacing phase until
the last block has a mobile displacing phase saturation.
Figure 3.7.9. Longitudinal pseudos.
Kro Krw
Swc Sw 1-Sorw
Measured Relative Permeability

Effective Relative Permeability
Figure 3.7.10. Horizontal integration.
This process can be extended using the notion of Buckley–

Leverett displacement so that the displacing phase is retarded until
the shock front saturation is attained in the aggregate cell. The
resulting relative permeability is shown in Fig. 3.7.11.
This leads to truncated relative permeability functions with very
steep transitions corresponding to the change in mobility at the shock
front. Such functions give severe problems for the Newton iteration
schemes of fully implicit models and, if used, should be modified to
have finite slopes.
Figure 3.7.11. Truncated relative permeability.
These techniques can be used to limit the effects of numerical

dispersion but can give numerical problems and also problems in
dealing with flow reversal because the choice of the outgoing face is
fixed.
3.7.5. Well Pseudo Relative Permeability

As well as describing the mobility of fluids in the reservoir as
a function of the saturation values in the neighbourhood, the
relative permeability functions relate the average saturation in a
cell connected to a production well with the produced fluids. The
behaviour of fluids flowing in the reservoir far from the well may not
in any way describe the interaction of the phases in the vicinity of
the well with the perforations.
A simple example of this is a well perforated over a small interval
in a thick cell which has water and oil in vertical equilibrium.
Until the water contact reaches the base of the perforations, no
water is produced and when the contact reaches the top of the
perforations, only water will be produced. Thus the linear behaviour
of the relative permeability functions is reflected but over a reduced
saturation interval compared to the function appropriate for cell-

to-cell calculations.
Well pseudo curves may be used to approximate the effects of
coning and cusping in coarse models which do not have sufficient
spatial resolution to accurately simulate these effects.
3.7.6. Summary
The use of relative permeability arises because of the requirement
to take into consideration the consequences of saturation dependent
mobility at the microscopic scale.
These functions, when used on a macroscopic scale in reservoir
models embody the microscopic effects compounded with assump-
tions regarding the distribution of reservoir properties and the spatial
distribution of fluids within the finite difference cell.
The effects of the spatial distribution of fluids and the vertical
distribution of horizontal permeability can be accounted for by using
upscaling techniques which are representative of the large finite
difference cells.
The grouping of cells is made by ensuring that the up-scaled func-
tions used in the coarse model adequately represents the behaviour
of the high resolution model.
3.8. Model Capillary Pressure

As described in Sec. 3.6, the capillary pressure is the difference
in pressure between two immiscible fluids and is a function of the
wetting phase saturation.
For a water-wet rock, the existence of a non-zero capillary
pressure implies that the rock can maintain some water above the
contact level where the water and oil phase pressures are equal.
The magnitude of the water saturation at any point depends upon
the balance between capillary and gravity forces.
The upwards capillary force is balanced by a downwards buoy-
ancy force arising from the density difference of the two phases and
the height above the free water level. For constant fluid density
difference, the saturation, at any height above the contact, at which
Figure 3.8.1. Model initialisation.
the capillary and gravity forces are balanced is a function of the

capillary pressure.
This gives rise to the transition zone of varying water saturation
observed in hydrocarbon reservoirs.
The technique used for initialising a reservoir model is shown
in Fig. 3.8.1. The procedure used is for each grid point (defined
by its cell centre) the downwards gravitational force, due to the
density difference times the height above the contact, is computed
and the corresponding equilibrium water saturation is obtained from
interpolation within the input capillary pressure data.
For cells which cut the contact there is the potential for errors
to occur as, if the cell centre is below the contact, the whole cell is
filled with water.
This can give rise to severe saturation errors at the edge of a

model with thick grid blocks.
In order to overcome this problem, some models provide the
facility of a more precise initialisation by subdividing the grid in
the vertical sense by some factor. Each subdivided cell is then used
as a calculation node for setting the saturation and the average
saturation in the cell is computed in a way which is more consistent
with the physical distribution of the fluids. This is known as the slice
integration technique.
This can give rise to non-equilibrium terms at initial conditions
as the inter block flow potentials are not computed using the same
method but simply using the cell centres to compute the gravitational
term and the average saturation for evaluating the capillary pressure
curve.
For any non-zero and nonlinear capillary pressure function, the
more precise initialisation will produce fluid flow potentials at initial
conditions.
The size of these terms is usually small compared to the
viscous pressure gradients set up by oil production but in special
circumstances (thick grid blocks and high capillary pressure) it is
advisable to check the movement by permitting some time steps prior
to start of production.
3.8.1. Manipulation of Capillary Pressure

In the same way that it was necessary to find a representative relative
permeability function for each rock type (see Sec. 3.7.1), the capillary
pressure data can be normalised to find a representative or average
curve.
The technique used is slightly different and is due to Leverett
who recognised that capillary pressure is a function of the surface
tension and the pore geometry. He used a relationship between the
mean pore radius

φ
r∝ (3.8.1)
k
giving for the capillary pressure

φ
pc (Sw ) = j(Sw )σ . (3.8.2)
k
The Leverett j curve then should be a function relating to a

particular rock type and differences in porosity and permeability
between the measured
sample and the model layer are accounted
for by the term φ/k.
This j function may be normalised over the dynamic range of the
water saturation using
∗ Sw − Swc
Sw = . (3.8.3)
1. − Swc
An example of such a curve is shown in Fig. 3.8.2.
Note that the water saturation range is different to the nor-
malisation for the relative permeability due to the difference in the
direction of movement of the wetting phase. Here we are dealing with
a drainage process whilst the relative permeability measurements
correspond to imbibition.
versus Sw
20
Average Curve
15
J (Sw)
10
0
0 20 40 60 80 100
Brine saturation percent
Figure 3.8.2. Typical leverett j curve.

For any model layer with porosity and permeability φi and ki the
j function can be used to compute the capillary pressure as:

Sw − Swc
i φi
pic (Sw ) = j ∗ σres . (3.8.4)
1. − Swc
i ki
The capillary pressure function obtained in this way may be used

in conjunction with the denormalised relative permeability functions
giving a consistent saturation function for each model layer.
3.8.2. Vertical Equilibrium

In the same way that the vertical equilibrium relative permeability
was developed in the previous section, we can derive the capillary
pressure function for a block, initially filled with oil, through which
the water contact rises.
Because the phase pressure difference is given at the finite
difference node (the cell centre), if we consider a rising contact with
no transition zone then for the contact at the base, middle and top
of the cell we obtain the three capillary pressures:
1
pcb = − ∆ρg∆z, (3.8.5)
2
pcm = 0, (3.8.6)
1
pct = + ∆ρg∆z. (3.8.7)
2
The capillary pressure function which results is shown in Fig. 3.8.3
and, like the corresponding relative permeability, is a linear function
of water saturation.
In order to obtain this relationship, we assumed that the
transition zone was negligible compared to the grid block thickness.
If this is not the case then we can compute pseudo capillary pressure
functions by integrating the water saturation through the transition
zone and relating this to the position of the water–oil contact.
As the contact moves then the phase pressure difference at the
cell centre and the average saturation can be evaluated, giving a
pseudo capillary pressure function. Because potentially this function
Figure 3.8.3. Vertical equilibrium pseudo capillary pressure.
is different for every cell, this approach, although giving the correct
oil in place and stability at initialisation, is little used and has been
superseded by the slice integration technique.
3.8.3. Summary
The concept of capillary pressure, as with relative permeability, arises
from the microscopic effects of immiscible fluids occupying the same
pore space. When used in a finite difference model, the capillary
pressure. Along with the relative permeability, acquires a component
which describes the spatial distribution of fluids inside the finite
difference cell.
The user, by selecting different input for these functions can
impose diffuse flow, segregated flow or whatever he believes most
closely reproduces the reservoir behaviour, and to a large extent
it is these assumptions and the resulting pseudo functions which
determine the results of the model.
3.9. Fluid Properties and Experiments

Oil and gas reservoirs contain complex mixtures of hydrocarbon com-
pounds together with non-hydrocarbon compounds at temperature
and pressure conditions which are completely outside the everyday
experience. The fluids are brought to the surface and processed
through separators so that physically and chemically stabilised,
non-toxic products can be sold. Some scientific approach to the

understanding of the behaviour of the fluids can therefore have an
impact on the quantity and quality of the sales stream, so that the
study of hydrocarbon fluid behaviour has been pursued with much
interest, leading to a large amount of data and theoretical models for
describing their properties.
3.9.1. Single Component Properties

Prior to embarking on a description of the properties of hydrocarbon
mixtures, it is useful to consider the simple case of a single component
fluid system such as water.
It is well known that the boiling point of water is a function of
the ambient pressure, and that this boiling point line in the pressure
and temperature (p−T ) plane has a limited extension, beyond which
no phase change occurs.
The line is called the boiling point line and the upper limit is
called the critical point. At this point, the intensive properties of the
liquid (those which are unrelated to the total mass) are identical to
those of the vapour.
We can arrange to cross the boiling point line from liquid
to vapour in many ways, but two simple ways are by fixing the
pressure and increasing the temperature or alternatively, by fixing
the temperature and decreasing the pressure.
The transition from one phase to another is not instantaneous, as
the increase in temperature or decrease in pressure can only continue
after all the liquid has vapourised.
At the boiling point line, although the temperature and pressure
are held constant for some time, the fluid volume is experiencing
extremely large changes.
3.9.2. Properties of Mixtures

If we extend the system so that we have two components, one of
which is more volatile than the other, then the region in the p − T
plane where there is liquid vapour equilibrium is extended out from
Figure 3.9.1. P –T Diagram for ethane N -heptane mixture.
the boiling point line to a two phase region. This is shown for a
binary system of ethane and N -heptane in Fig. 3.9.1.
This two phase region also has limits in both temperature
and pressure, but unlike the single component where the limits in
temperature and pressure correspond to the critical point, the limits
for a mixture are not coincident with the critical point.
As may be seen in Fig. 3.9.1, the behaviour of a mixture of two
components is not a simple linear interpolation of properties and the
critical pressure for the mixture may be significantly greater than
that of either component.
As we cross the two-phase region from liquid to vapour, at

constant temperature say, then the liquid saturation and the fraction
of the volatile component in the liquid both decrease. The lines of
equal liquid saturation (called isovols) or lines of equal composition
(quality lines) can be drawn onto the two-phase region to give a
quantitative expression of the change in the phase equilibrium across
this region.
3.9.3. Hydrocarbon Types

The p–T diagram is a useful guide to the different types of hydro-
carbons encountered. It should be recognised that the behaviour of
a hydrocarbon mixture depends not only on the composition of the
mixture but also on the temperature and pressure of its environment.
This is illustrated in Fig. 3.9.2.
For example, a mixture which is a volatile oil at one temperature
and pressure, would become a two-phase oil and gas system at
lower temperature and pressure or a retrograde condensate at higher
temperature and pressure.
Figure 3.9.2. Effect of different fluids and temperature.

The classification of different hydrocarbon types is usually made

on the gas–oil ratio, and this is related to the behaviour in the p–T
plane as both temperature and pressure increase with increasing
depth.
black oil GOR < 1,000 scf/STB
high shrinkage oil GOR < 3,000 scf/STB
volatile oil GOR < 5,000 scf/STB
retrograde condensate GOR < 30,000 scf/STB
wet gas GOR < 100,000 scf/STB
dry gas GOR > 100,000 scf/STB
The corresponding pressure versus temperature diagrams are shown
in Fig. 3.9.3.
3.9.4. Definitions
Prior to discussing the experimental methods used to characterise
hydrocarbon systems, we will define some of the more important
terms used.
Saturation pressure
This is the pressure at which, during compression or expansion
at constant temperature, the gas system is in equilibrium with an
infinitesimal quantity of liquid or the liquid system is in equilibrium
with an infinitesimal quantity of gas.
Bubble point pressure

For a specific temperature this is the saturation pressure where the
liquid is in equilibrium with an infinitesimal quantity of gas.
Dew point pressure

For a specific temperature this is the pressure at which the gas is in
equilibrium with an infinitesimal quantity of liquid.
Oil formation volume factor

The oil formation volume factor is the ratio of the volume in the
reservoir at high temperature and pressure to the volume of the oil at
P P
Low shrinkage High shrinkage
T T
P P
Volatile Condensate
T T
P P
Wet gas Dry gas
T T
Figure 3.9.3. Fluid types.
standard conditions which results from producing the hydrocarbons

to surface and passing them through the separation facilities.
Gas formation volume factor

This is the ratio of the volume of gas in the reservoir to the volume
of the same mass of gas at standard conditions.
Gas deviation factor

The gas deviation factor is a factor used to modify the ideal gas law
to take account of the finite size of and the interactions between gas
molecules. For a given temperature, this relates the pressure volume
product to the temperature value
pV = znRT. (3.9.1)
The gas deviation factor and the gas formation volume factor are
related by

ps T
Bg = z, (3.9.2)
pTs
where ps and Ts are the pressure and temperature at standard
conditions.
Solution gas content

This is the quantity of gas which can be dissolved in the oil at that
specific temperature and pressure.
Oil gravity
This is the ratio of the oil density to the density of water. This is
also quoted in terms of API gravity where:

141.5ρwater
γAPI = − 131.5. (3.9.3)
ρoil
Gas gravity
The gas gravity is the ratio of the density of the gas to the density of
air at standard conditions. At low pressure, the deviation factor for
both gases is approximately unity and so the gas gravity is given by
Mg
γg = (3.9.4)
Mair
where M is the molecular weight.
Coefficient of thermal expansion

This reflects the expansion of the fluid as a function of temperature.
1 dV
cTemp = . (3.9.5)
V dT
Isothermal compressibility
The isothermal compressibility is the relative change in volume per
unit pressure change:
1 dV
coil = − . (3.9.6)
V dp
The following experiments set out to determine values for these
quantities which specify to some reasonable accuracy, the behaviour
of the hydrocarbon system.
3.9.5. Experiments
For a black oil system, the experiments are designed to provide a
means of evaluating the pressure dependent fluid properties required
in the formulation of models from material balance to fully implicit
numerical models.
3.9.5.1. Constant composition expansion

The schematic experimental sequence is depicted in Fig. 3.9.4 where
it can be seen that the sample is recombined at high pressure, and
the temperature is raised to the reservoir temperature. During this
phase, the coefficient of thermal expansion is measured. At a pressure
above the reservoir pressure, the initial oil volume is measured and
the pressure is reduced by increasing the cell volume by withdrawal of
mercury. The oil volume is measured and the point at which bubbles
evolve is noted.
The pressure is continuously decreased, keeping the evolved gas in
contact with the oil. In some experiments, the oil volume is reported
through the pressure drop cycle. The pressure is decreased until
standard pressure is attained and the temperature is reduced to the
standard value. The total amount of gas liberated and the gas gravity
is recorded as well as the volume of the oil. The recordings taken
during this experiment are indicated in Fig. 3.9.5.
3.9.5.2. Differential liberation

In the differential liberation experiment, the initial part is the same
as for the constant composition expansion. The coefficient of thermal
Figure 3.9.4. Constant composition expansion.
Figure 3.9.5. CCE results.

expansion and the fluid compressibility above saturation pressure are

measured.
The procedure below the saturation pressure is to decrease the
pressure in the cell from the saturation pressure to atmospheric
pressure in 10 equal steps. At each step, the oil and gas system are
allowed to come to equilibrium and all of the gas is displaced out of
the cell. The volume of oil and gas are measured and the gas density
is measured. This is illustrated in Fig. 3.9.6.
This process is repeated for each pressure in the series, so that
at each pressure reading, the cumulative gas evolved and its density
can be calculated. This experiment then gives volumes of the oil and
Figure 3.9.6. Differential liberation.

Figure 3.9.7. Differential liberation results.
gas below the bubble point as the pressure decreases. This is shown
in Fig. 3.9.7.
In Sec. 3.9.2, we saw that the behaviour of a binary system
in terms of the component properties was not at all linear. By
performing the differential liberation experiment, at each pressure
step we create a new mixture and it would be unlikely that the
combined effect of the 10 pressure decrements would give the same
result as the constant composition expansion.
In fact, the total amount of gas released (both mass and volume)
tends to be larger for the differential liberation compared to the
constant composition expansion.
3.9.5.3. Constant volume depletion

The constant volume depletion experiment consists of the initial part
as for the constant composition expansion and differential liberation.
Below the bubble point, the pressure is decreased in a way similar
to the differential liberation except that at every step, the gas is
displaced until the initial cell volume is attained.
Thus the next pressure decrement is performed with the oil in
contact with some of the gas from the previous steps, which alters
the equilibrium compositions for the oil and gas.
The measured quantities for this experiment are the same as for
the differential liberation but the quantity of gas released is usually
smaller.
3.9.5.4. Separator tests

In order to better understand the relationship between the reservoir
fluid and the sales stream, the reservoir fluid is subjected to simulated
separation facilities.
This might consist of two pressure decrements and a final
drop to standard conditions. During the pressure decrements, the
temperature is also reduced so that the volume of gas evolved is
minimised.
The tests will normally be performed for a range of separator
conditions so that the optimum values in terms of retention of
hydrocarbons in the liquid phase can be estimated.
3.10. Model Fluid Properties

From the laboratory data obtained, it is necessary to extract the
information required to perform engineering calculations. The data
requirements depend on the formulation of the calculations and below
we describe the form of the data and the manipulation processes for
the simplified case of a black oil model.
3.10.1. Black Oil Fluid Properties

Prior to discussing the black oil assumptions and their representation
in a finite difference model, we will define some terms which are
generally used in reservoir simulation models for black oil or more

complex systems.
Component
This is a constituent of the hydrocarbon mixture which might
correspond to a particular chemical species, a group of hydrocarbons
within a range of carbon numbers or a statistical mixture of com-
pounds evaluated by some means.
Phase
This refers to the state of a group of components which at the defined
temperature and pressure form a coherent mixture with measurable
properties. Thus in a saturated reservoir, the gas cap gas is an
example of a phase.
Phase mole fraction of a component

This is the ratio of the number of moles of that component compared
to the total number of moles of all components existing in that phase.
As a convention, the liquid mole fraction of a component i is
written as xi and the gas mole fraction as yi.
Phase equilibrium
This is the state of dynamic equilibrium achieved when at a
stable temperature and pressure, the compositions of the phases are
independent of time, that is all values of x and y are constant.
At equilibrium, we can define a relationship between the x and
y values for the various components as
yi = Ki xi , (3.10.1)
where
Ki = K(T, p, xi , xj , . . .) (3.10.2)
and K is called the equilibrium coefficient.
The computation of liquid vapour equilibrium is mathematically
difficult and in simulation of complex processes can consume a large
fraction of the computer time.
Black oil
In a black oil model, it is assumed that the hydrocarbon system can
be adequately represented by two components with a distribution
of molecular weights. This distribution is that determined by the
separation facilities so that the two components correspond to stock
tank oil and gas (separator gas plus the gas evolved from the
separator liquid in going to tank condition).
The equilibrium coefficient for these two components at a fixed
reservoir temperature is assumed to be a function of pressure only.
In addition, all reservoir processes are assumed to be isothermal
and the temperature effects in producing to surface are accounted for
in the formation volume factor which relates surface to subsurface
volumes.
An assumption which is usually made (but can be relaxed) is
that the stock tank oil is non-volatile even at reservoir conditions
and that it does not partition into the vapour phase.
This implies that the vapour mole fraction for the gas component
is unity and that the reservoir gas has a fixed composition identical
to the surface gas.
Together with the assumption regarding the equilibrium
coefficient
ygas = K(p)xgas (3.10.3)
we see that the ratio of the number of moles of stock tank oil in
the liquid phase to the number of moles of gas in the liquid phase is
determined uniquely by the equilibrium coefficient, K(p).
This means that for a given mixture of components at a set
pressure, knowing the form of K(p) (which is a transformation of
the solution gas content), the liquid vapour equilibrium mixture can
be computed a priori with no iterations.
Water is assumed not to partition into either hydrocarbon phases
and not to influence any of the hydrocarbon properties.
The data required for a black oil model is then
(1) the oil formation volume factor versus pressure,

(2) the solution gas–oil ratio versus pressure,
(3) the oil viscosity versus pressure,

(4) the stock tank oil gravity,
(5) the gas formation volume factor versus pressure,
(6) the gas viscosity versus pressure,
(7) the gas gravity,
(8) the initial water formation volume factor and
(9) the water compressibility.
3.10.2. Data Manipulation

If the simulation is expected to enter the two (hydrocarbon) phase
region, then the pressure dependence of the solution gas–oil ratio is
required.
The experiment which gives estimates for this is the differential
liberation experiment but this also tends to over estimate the
total amount of gas liberated. (It accurately simulates a 10 stage
separation at reservoir temperature).
The accepted approach is to use these data scaled to give the
correct total gas–oil ratio and to assume that the form of the curve
between the bubble point and the stock tank conditions scales in the
same way.
The simple linear scale factor is defined as

Rsf (pb ) Rd (pb ) − Rsd (p)
Rs∗ (p) = f
− s d
f
Bob . (3.10.4)
Bob Bob
Here superscripts refer to either flash or differential liberation and

the subscript b indicates that these values correspond to the bubble
point.
For consistency, this equation has been written using Rs as the
volume of gas dissolved in the oil at that pressure. Laboratories
frequently report the value which they measure which is the volume
of gas (R∗ ) evolved from the oil. For the differential liberation
experiment, this corresponds to the term
Rs∗ (p) = Rsd (pb ) − Rsd (p). (3.10.5)

The formation volume factor is also scaled so that the bubble point
value corresponds to the flash experiment.
f
Bob
Bo (p) = Bod (p) d
. (3.10.6)
Bob
Here, the correct values have been taken as those from the constant
composition expansion (flash) experiment. If separator tests have
been made it is preferable to use those which most closely resemble
the field conditions.
After performing the scaling, we have corrected tables of data,
which for oil properties typically appear as in Figs. 3.10.1–3.10.3.
Figure 3.10.1. Oil formation volume factor.
Figure 3.10.2. Solution gas–oil ratio.

Figure 3.10.3. Oil viscosity.
These are however not in the form required by reservoir simulation

models.
This is because of the requirement to treat the bubble point as a
time dependent variable.
Consider an oil reservoir, which initially contains oil having the
same properties everywhere, whose pressure falls below the bubble
point and then rises again. Due to the different mobility and density
of the liquid and vapour phases, some segregation might occur during
the period of reduced pressure.
When the pressure rises, the pressure dependent equilibrium
inherent in the model will cause the gas in the vapour phase to
partition into the liquid phase raising the saturation pressure. This
process ends either when all the gas is in the liquid phase or when
the liquid phase is saturated.
For those cells at the base of the reservoir model which are
deficient in gas the first case will occur and the model must compute
under saturated values corresponding to a bubble point lower than
the initial value.
If the pressure continues to rise, the gas rich cells will also
reach saturation pressure and pass into the under-saturated region,
requiring the model to compute first the saturated quantities and
then the under-saturated quantities at bubble point pressures higher
than the initial value.
The model then requires data which enable it to compute oil

properties not only along the line defined by the modified differential
liberation data, but for any value of bubble point pressure and
reservoir pressure.
For this reason, the saturated solution gas–oil ratio and formation
volume factor must be entered, extending above the expected
maximum pressure.
The formation volume factor in the under saturated region is then
calculated using a value of the oil compressibility or by interpolation
between several sets of under-saturated data.
The treatment of the under-saturated viscosity is similar imply-
ing a requirement for a viscosity compressibility factor. The oil
compressibility (if used) and viscosity compressibility are considered
as constant.
Using a pressure p in the under saturated region, we can define
the compressibilities
1 Bob − Bo (p)
co = , (3.10.7)
Bob p − pb
1 µo (p) − µob
cµo = . (3.10.8)
µob p − pb
For any new value of bubble point pressure, p∗b , then we can write
Bo (p) = Bo (p∗ )(1 − co (p − p∗ )), (3.10.9)

µo (p) = µo (p∗ )(1 − cµo (p − p∗ )). (3.10.10)
Such a treatment of fluid properties allows the bubble point to

become a variable in the simulator.
3.10.3. Spatial Variations

The assumptions used to develop the black oil description appear to
be extremely restrictive. Some flexibility is available in attempting to
model spatial variations of fluid properties, but the two component
formulation together with the requirement of equilibrium at initial
conditions restrict such variations to be depth related.
The reason for this is that variation of fluid properties in a way

not related to depth would lead to different fluid densities at the
same depth, so that even if at that depth all fluid potentials were
zero, at depths above and below, non-zero potentials would exist.
3.10.3.1. Variable bubble point

The initial composition of the oil is frequently a function of depth.
Typically the bubble point pressure decreases with depth, which
together with the increasing pressure gives a slight curvature to the
pressure versus depth curve.
Because the pressure at a given depth depends on the integral of
the fluid densities above it, the pressure is calculated on a regular and
relatively fine vertical grid so that the system of equations converges
rapidly.
3.10.3.2. Variable Api gravity

In a similar way, the density of the stock tank oil may increase with
depth giving rise to a similar nonlinear pressure versus depth curve.
In both cases, the description of the spatial variation is made
using an initially different mixture of components to describe the
fluids. As long as the components are treated in the mass balance
equations, then the initial variation will be reflected in the subsequent
movement and production characteristics.
3.11. Aquifer Treatment

A primary source of reservoir energy for hydrocarbon production
is the expansion of the water zone in contact with the reservoir.
Because of the potentially large volume of water involved compared
to the reservoir volume this expansion can be extremely important
for the reservoir production characteristics.
In principle, the aquifer could be included in the reservoir model
as part of the finite difference grid (see Fig. 3.11.1), but to do so
would entail a large overhead in computing water flux within the
(poorly defined) aquifer region when the main concern is only how
much water flows into the reservoir during some time interval.
Figure 3.11.1. Aquifer in grid.
Various techniques have evolved to enable a realistic but efficient

simulation of the aquifer response.
3.11.1. Hurst Van Everdingen

Hurst and van Everdingen used a modification of the solution of the
radial form of Darcy’s law used for well test analysis, solving for the
rate by specifying a constant terminal pressure.
They then evaluated dimensionless water influx functions for
infinite radial and linear aquifers.
For a constant terminal pressure, the cumulative water influx was
given by
We = U ∆pWD (tD ), (3.11.1)
where
We = cumulative water influx,
U = aquifer influx constant,
WD = dimensionless water influx function,
tD = dimensionless time.
The cumulative water influx, after n + 1 time steps, for a cell coupled
to the aquifer is

j=n
We (tn + 1) = U ∆pj WD (tDn+1 − tDj ). (3.11.2)
j=0
This requires that for each cell coupled to the aquifer, the history of
pressure changes must be stored and used to compute the current
value of the influx.
Because of the computational overhead involved, more approxi-
mate methods have been developed.
3.11.2. Carter Tracy

Carter and Tracy approximated the water influx as a set of intervals
with constant rate of water influx. The approximate difference in
cumulative influx was written
We (tn+1 ) − We (tn ) = (tDn+1 − tDn )

 

 u∆pn+1 − We (tn ) dp 
dtD n+1 
D
, (3.11.3)

 pD (tDn+1 ) − tDn dp D 

dt D n+1
where
∆pn+1 = total pressure drop,
pD (tD ) = dimensionless pressure.
The dimensionless pressure function is the constant terminal rate
solution to the radial form of the diffusivity equation.
3.11.3. Fetkovitch
Fetkovitch approximated the rate of influx into the reservoir using
an analogy of the productivity index
dWe
= J(p − pn+1 ), (3.11.4)
dt
where
J = the Fetkovitch influx constant,
p = the average aquifer pressure,
pn+1 = the pressure of the cell.
For finite aquifers, the influx into the reservoir is taken into account
to compute the average pressure by material balance for the next
time step.
3.11.4. Numerical Aquifer

In order to try to model the transient pressure response of the aquifer
without the need for a large number of cells, some models permit the
user to define an auxiliary coarse 1D aquifer model which is coupled
(using special transmissibility connections) to the main model. This
is illustrated in Fig. 3.11.2.
This approach overcomes the problems of the approximations
used for the analytical influx functions, but does leave the potential
for including large space truncation errors into the model by coupling
it to an extremely large grid block in the aquifer model.
To avoid this, the size of the grid blocks in the aquifer should be
graded upwards in a smooth way from dimensions similar to those
used in the reservoir zone.
Figure 3.11.2. Numerical aquifer.

3.12. Model Well and Production Data

Up to this point, with the exception of well pseudos, we have been
concerned with modelling the properties of the fluid flow in the
reservoir. In fact, this flow is initiated and sustained by production
from wells and injection into wells and a realistic treatment of the
behaviour in and around the well is a necessary step in the process
of reservoir simulation.
3.12.1. Well Inflow

For a well located in some drainage area with a radius of re and
a pressure at the external radius of pe , the operationally derived
relationship for the flow rate of oil from the formation is
qo = P I(pe − pwf ). (3.12.1)
From the solution of Darcy’s equation, the productivity index, P I,

can be expressed as
2πkkro h
PI = (3.12.2)
µBo ln rrwe + S − 0.5
for the case of semi steady state, and
2πkkro h
PI = (3.12.3)
µBo ln rrwe + S
for the steady state solution.

These become
2πkkro h
P I∗ = (3.12.4)
for the case of semi-steady state and
2πkkro h
P I∗ = (3.12.5)
for steady state when the productivity index is expressed in terms of

the average pressure, p
qo = P I ∗ (p − pwf ) (3.12.6)
From these equations we can see that the productivity index is

dependent on the fluid saturation in the neighbourhood of the
perforations, via the relative permeability term, and also dependent
on the pressure via the viscosity and formation volume factor.
Further, the index which controls the flow of fluid from the
formation into the wellbore depends on the distance from the well at
which we reference the reservoir pressure.
For a well in a reservoir model, these effects are accounted for in
the following manner.
The inflow to the well from a cell at a pressure p and water
saturation Sw is written
qo (p, Sw ) = Twell Mo (p, Sw )(p − pwf − ε), (3.12.7)
where
qo = the volumetric flow rate of oil,
Twell = the well connection factor,
Mo (p, Sw ) = the oil mobility term,
pwf = the flowing bottom hole pressure,
ε = the gravity correction term.
The well connection factor (or well index) is defined as
αkh
Twell = , (3.12.8)
ln rrwo + S
where
α = the unit conversion factor,
kh = the permeability thickness,
ro = the pressure equivalent radius,
rw = the wellbore radius,
S = the skin factor.
Here, the pressure equivalent radius is usually obtained from the

publications by Peaceman, the current one being:
"# #
ky 2+ kx 2
kx δx ky δy
ro = 0.28 "# "# . (3.12.9)
ky kx
kx + ky
The mobility term, M (p), contains the saturation and pressure

dependent terms leaving the well connection factor as a constant
for a given perforation. This mobility term is expressed as
kro (Sw )
M (p, Sw ) = . (3.12.10)
µo (p)Bo (p)
For a high productivity index well with a low draw down, the
computed flow rate may be sensitive to the gravity correction, ε,
which corrects from the well reference depth to the sand face depth.
Various methods are available for this calculation depending on
the treatment of the multi-phase density and the dating of the
calculations. For very high deliverability wells, the results should be
checked for sensitivity to this correction.
3.12.2. Production Control Data

The reservoir engineer performing a simulation study is faced with
the task of representing the observed production data to the simu-
lator, in order to try to match the reservoir behaviour, and then to
attempt to predict and optimise the likely future performance of the
reservoir taking account of the facilities and operational constraints.
As each field is unique, the task of providing the engineer with a
sensible but manageable choice of options has been one of the more
difficult problems of practical reservoir simulation.
3.12.2.1. Targets
Target rates are those rates which the simulator should attempt to
maintain provided that this can be achieved within the set operating
constraints.
Rates may be specified in a variety of ways for different purposes.

Examples are surface oil rate, gas rate, water rate, liquid rate or
subsurface oil rate or total fluid rate. For injection wells, additional
target rates can be given expressing the required rate as a function
of the phase production (for example reinjecting 85% of produced
gas) or reservoir voidage (for example total injection rate balancing
the voidage rate) to give, to first order, a stable average reservoir
pressure.
Inspection of the productivity index equations above, reveals
that they are written for the oil phase. Similar expressions exist
for the water and gas phases (although the gas equation may be
more complex due to rate dependent skin factor and the difference
in pressure behaviour of gases leading to the equation being linear
in the real gas pseudo pressure).
This implies that if the user requests a well to produce a certain
rate of oil per day, then whatever the ratio of the mobility terms for
the water and oil phases, the well will produce fluids to give that
water–oil ratio unless some constraint is activated.
This might be necessary as the well might not be capable of
flowing at the simulated bottom hole pressure and water cut.
3.12.2.2. Constraints
Because a well in the simulation model could produce fluids in a
way impossible for a real well, the user can elect to choose some
constraints which can be applied.
These constraints may be related to wells, (for example a
minimum tubing head pressure for all wells corresponding to the
pressure needed to deliver hydrocarbons to the separator) or to
groups of wells (a group of wells producing into one separator has
a maximum group liquid rate determined by the separator size and
retention time).
The constraints can be connected in a tree structure so that at
each node (well, platform and field) constraints are honoured.
As well as gas–oil ratio limits and water cut limits, secondary
rate limits can be set which will permit the primary target rate to
be met provided that the secondary limit is not violated. This can
be useful where there are maximum through put restrictions such as

produced water disposal or gas treatment.
A useful constraint to activate is the economic limit which shuts
in wells or terminates the simulation when lower limits for production
rates are crossed.
In terms of oil field operation, if a well has some production
problem then frequently the production is made up either by working
over the existing well or by drilling a new well.
Reservoir models have a user definable set of actions which can
be performed automatically in the model to attempt to simulate
reasonable oil field practice in maintaining production levels.
3.12.2.3. Actions
Typically these actions consist of attempting to control the produc-
tion of unwanted components such as water and gas by either altering
the well rate (reducing the rate of high water cut wells), performing
a work over on the well to squeeze off the worst offending perforation
(this has a 100% success rate in models) or the well can be shut in.
If a group of wells has spare production capacity, then at some
predefined drilling rate, wells can be added in specified locations in
order to increase the production rate to correspond to the available
facilities.
3.12.3. Practical Considerations

Full field reservoir simulation, which is relevant for the constraints
and actions discussed above, is often performed in the stages
corresponding to a refinement of reservoir description by matching
the observed behaviour followed by a series of prediction runs which
attempt to optimise the reservoir performance in some way.
These two modes of operation are very different in the way in
which target rates and constraints are specified.
3.12.3.1. History match
During the period of a history match, oil gas and water rates will
usually be available for each well on some reasonably frequent time
basis.
In addition, down-hole reservoir pressure data at specific well

locations will also be available.
The engineer performing the simulation study has to attempt to
match the pressure behaviour and water movement in the reservoir
from the well control data at his disposal.
If the input data to the model is specified in terms of produced
oil rate, then for wells where the water cut is poorly simulated, the
total volumetric production from the reservoir may be seriously in
error giving rise to spurious pressure behaviour.
Also, the inclusion of (sensible) well constraints can have adverse
consequences as for a well where the water cut is too high in the
model, the well will be shut in where as it continued in production
in the field.
Thus the rate specification and the inclusion of well constraints
can lead to the incorrect volumetric fluid withdrawal making the task
of achieving a history match almost impossible.
The problem becomes significantly more tractable if the well
constraints are deactivated and the production rate data is converted
to the equivalent volume of reservoir fluid at the ambient pressure,
and this rate is produced regardless of the proportion of each
component.
In approaching the simulation in this way, the problems of
matching the pressure and water cut behaviour or gas–oil ratio
behaviour are largely decoupled.
3.12.3.2. Prediction
During the prediction phase it is necessary that the production
constraints such as minimum tubing head pressure and maximum
liquid through put are implemented.
Figure 3.12.1 shows typical tubing lift curves for different water
oil ratios. If several layers are perforated each having different water
saturation, changing the bottom hole pressure can modify the water
cut and the convergent bottom hole (and therefore tubing head)
pressure must be found by an iterative procedure.
As in most iterative schemes, the closer the start value is to the
solution, the faster the convergence and so during a prediction it
Figure 3.12.1. Tubing lift curves.
is worth while setting target rates which are close to the expected
production rate.
It is a measure of the quality of the history match that the
model should proceed from the history match to the prediction phase
without experiencing instability or discontinuity.
Chapter 4
History Matching
Deryck Bond
Senior Consultant for Kuwait Oil Company
4.1. Introduction
It may be appropriate to start by attempting to define “history
matching” in the context of reservoir modelling. One possible
definition, variants of which are commonly used, is as follows:
“History matching is the process of modifying the model input data
until a reasonable comparison is made with historical data.”
In this definition the emphasis is on producing a model where

there is agreement between simulated and historical data. An
alternative definition that may be preferable is
“History matching is the process of making reasonable changes to
model input parameters to better match historical data and to
improve the predictive capability of the model.”
This latter gives emphasis to the business aim of our work:

producing a model that gives reasonable predictions of future
performance that can influence our business decisions.
Either definition raises a number of questions, such as:
• How well do we need to match historical data?
• What changes to our original model are reasonable?
• How do we go about making suitable changes?
• Are there multiple ways we can match the observed data?
209
• How are our conclusions influenced by the limitations of the

simulation tools we have?
• How can we form a view on the predictive capabilities of our model?
• Will a single (history-matched) mode allow us to meet our
objective or do we need to develop multiple models?
• What if we cannot achieve a reasonable match? Has an exercise
that has just invalidated a model been of any value?
These questions are common to modelling many “inverse
problems” — problems where we wish to match model “outputs”
(which may have significant uncertainty) by changing inputs (about
which we have limited information). Other inverse problems in the
petroleum industry are encountered in pressure transient analysis
and in seismic interpretation.
History matching reservoir models is a relatively complex inverse
problem. The underlying equations are non-linear; the tools used
to solve these equations may have limited accuracy and the physical
system is complex. Finding solutions to these types of problems is an
active area of mathematical research. For any inverse problems there
are a number of concerns related to how “well-posed” the problem
is. These include the following:
(1) Is there a solution?
Given that we are attempting to model historical data it would
seem reasonable that there would be a solution. However there
may be concerns about the limitations of the simulation model; if
the simulation model has insufficient numerical resolution then
we could envisage it not being possible to match the observed
date with the simulation model.
(2) Is the solution unique?
We shall see through examples that this is a problem that has
practical implications for us. If we can match observations in
many ways, then how much confidence do we have in our ability
to predict?
(3) How “smoothly” should we model outputs depends on
model inputs?
This has clear implications for us being able to match the model.
History Matching 211
Before attempting to address the above questions, we will put the

process of history matching into an appropriate context. Firstly, it
is necessary to put the exercise (and the wider reservoir modelling
exercise that it is part of) into an appropriate business context;
secondly, it is important to have a view of the history matching
process as a part of the wider reservoir modelling work flow.
This is followed by a description of the work that would be
involved in a “conventional” history matching exercise involving
matching a single model of a reservoir and its use for prediction.
This type of approach is summarised in the literature of Mattax
and Dalton (1990). In doing this, emphasis is placed on some
of the “practical” elements of the study, particularly on issues
related to surveillance data, data acquisition, and quality control
(QC) and “calibration” of the matched model to field performance.
A process of matching the data is described that could use “manual”
or “computer-assisted” approaches to altering model parameters.
Problems associated with non-uniqueness are discussed.
Finally, comments and some references to automatic history
matching and to the use of multiple models for uncertainty assess-
ment are provided. These notes do not do justice to these very
important issues.
4.2. Context of the History Match Study

As for any study, it is important to understand the role that the
history match study plays in the wider context of achieving business
goals, its relation to choices about reservoir management, and its
relationship to the wider range of technical work that would be
involved in a geological and simulation study.
4.2.1. The Business Context of the Study

For anyone involved in a history match exercise, having (and keeping
throughout the study) a clear understanding of the business context
of the study is vital. This is because the criteria for judging the
“fitness for purpose” of the matched model will be a function of the
business aims. Some examples can illustrate this.
Example 1 — A study of potential in-fill locations in a large mature

field with a large volume of data. In this case, success would be
judged by the ability to guide the choice of well location. The match
to production and surveillance data, especially close to any proposed
in-fill locations, would be expected to be good.
Example 2 — A study early in the life of a field intended to identify

the time when artificial lift may be required. In this case success
would depend on the ability of the model to predict factors that
would influence the need for artificial lift (pressure decline and the
tendency of well water cuts to rise). Achieving a good match to
individual well data may be less important than in Example 1. On
the other hand, assessing uncertainty may be more important; the
business decision could be based on an estimate of the earliest that
artificial lift would be needed.
Example 3 — A study aimed at predicting short- to medium-term

production. This may be important to financial planning. In this
type of application there would be a particular interest in ensuring
that well performance was well matched at the start of the prediction
period.
The scope for having multiple models in the above cases differs.
In the first example the appropriate style of model (lots of history and
lots of data) may be relatively deterministic. This, and the increased
efforts needed to history match a field with a lot of data, could make
having a single model a sensible choice. For the second example, the
emphasis on assessing uncertainty and the reduced effort that may
be associated with matching models with less data may make using
multiple models both more useful and more practical.
4.2.2. Relation to Reservoir

Development/Management
The reservoir studies of which history matching forms a part
cannot be viewed in isolation from the wider subject of reservoir
management. The ability to develop history-matched models that are
useful in reservoir management may be limited by decisions that are

made relating to how to develop the reservoir, what data to acquire,
and how to analyse and manage those data.
The following can reduce our ability to produce a useful history
match:
• Choosing to have commingled production from different “flow
zones” (or even different reservoirs) without making efforts (such as
production logging or using down hole measurements) to measure
such production.
• Having wells that are in poor mechanical condition and can allow
cross-flow between different sets of perforated intervals and/or flow
behind pipe.
• Having limited reservoir characterisation and surveillance data (for
example limited core data, the choice not to have 4D seismic
data, infrequent bottom hole pressure measurements, limited fluid
sampling, etc.).
• Having poor production (or injection) allocation because wells are
not tested frequently enough.
• Not running Repeat Formation Tester (RFT) (RFT — a Schlum-
berger trademark — or similar tool) or similar logs on development
wells.
• Failing to invest in good data management.
In all the above, there is a choice between generally lower- and higher-
cost approaches to managing a reservoir. (Coring wells costs money;
commingled production can save a lot of money.)
It is important in the context of a history match that the role
of such decisions is recognised. Expectations about the benefits of a
history match exercise will depend on previous reservoir management
decisions.
4.2.3. The Work Flow Context of the Study

At this point, it is as useful to provide a very high-level view
for history matching. The following flow diagram attempts to
illustrate this.
Data /Studies Models
Structural model
Seismic data
Wellbore seismic
4D seismic
Core data RCAL/SCAL Geological /
Sedimentalogical studies petrophysical
Fault seal studies geocellular
Open hole log data model
Petrophysical studies/studies of contacts/FWLs
Geochemical data Up-scaled geo-
RFT data model
Pressure transient data
Cased hole logs (PLT/TDT)
ProducƟon data IniƟal simulaƟon
StaƟc surveys/SIBHP data model
Flowing surveys/FBHP data
Material balance/Aquifer characterisaƟon studies
Part reservoir simulaƟon studies
History matched
Pressure maps
simulaƟon model
Fluid movement maps
Studies of grid size/pseudo relaƟve permeability studies
CompleƟon integrity studies
Dynamic model
for predicƟons
Figure 4.2.1. Overview of a typical reservoir modelling exercise.
Figure 4.2.1 shows the sequence of models that would normally be

produced in the course of a reservoir modelling exercise. Also shown
is an (incomplete) list of data and supporting activities that would
be used in the modelling work. An attempt has been made to relate
these activities to one of the models, the structural model. As can be
seen, a wide range of data and studies contributes to the structural
model, including both “static” and “dynamic” data, as discussed in
the next section. It may be instructive to attempt to do this for the
other modelling stages.
Figure 4.2.1 also shows the styles of iteration that may be
associated with the history matching process.
Firstly, the initial simulation model may be “edited”. This
is entirely appropriate for some types of changes that would be
introduced during the history matching process; examples include
making changes to estimated rock compressibility, making global

changes to permeability, and changing fault transmissibilities. For
other cases, for example changing the style of permeability variation,
it may be an acceptable way of exploring how a better match could
be achieved but would be unsatisfactory as a final model.
Secondly, we may wish to update the geocellular model in a
more systematic way than simply editing the simulation model input
parameters. This could involve generating a different realisation of
the geological model, exploring a different permeability model, or
introducing a greater degree of “determinism” into the geological
model.
Thirdly, we may wish to rebuild the structural model. This
may be done in order to better model connectivity in the reservoir.
There is currently less tendency to change the structure than other
elements of the geological model during history matching. This is not
a reflection of structural uncertainties being less important than, for
example, uncertainties related to the sediment logical model; it is a
reflection of the time-consuming nature of the structural modelling.
In all of the modelling stages illustrated in Fig. 4.2.1, consid-
eration needs to be given to issues related to gridding. These are
discussed later.
4.3. Static and Dynamic Data/Static and

Dynamic Models
4.3.1. Static and Dynamic Data
There is a tendency to classify data as being “static” or “dynamic”
and also to relate these data to the static (structural/geological) and
dynamic (reservoir simulation) models that we build. It is not always
straightforward to classify data in this manner.
Examples of “static data” include core descriptions and poroperm
measurements on core data. Examples of dynamic data include
production data and measurements of flowing and shut-in pressures
in wells. Some data, for instance wire-line log data that can indicate
water movement, would not fit easily into either category.
4.3.2. Dynamic Data and the Static Model

For a producing reservoir there may be a lot of dynamic data
available when the static geological model is being prepared. The
extent to which these data are taken into account when constructing
the static model will have a clear impact on subsequent history
matching. It is generally considered good practice to account for some
of these data in the static model. This is done with the hope that
the process of history matching will be easier and that the resulting
history-matched model will be more useful.
An example is, using pressure transient test data to help in
producing geologically based models for model permeability. Using
this approach would make it likely that well productivity would be
relatively well-matched in the simulation model prior to making any
adjustments to the permeability. Changes made during the process
of history matching may then be relatively modest and should not
detract from the geological plausibility of the permeability model.
An alternative would be to build the initial model for permeabil-
ity based solely on estimates from core. Such a model may require
far greater modification to match historical data.
The following are examples of how dynamic data could be used
in construction of the static model:
• use of pressure transient data, production log data and well

productivity data to constrain the permeability model;
• use of RFT data and data on fluid movement from well logs to
provide relatively deterministic models of the extent of baffles or
barriers;
• use of pressure transient data/interference test data to help refine
the structural model — fault locations and initial estimates of
transmissibility of faults;
• use of geochemical data or pressure maps to better understand
reservoir compartmentalisation;
• use of maps/cross-sections/fence diagrams illustrating fluid move-
ment (produced by production geology studies) to help better
define the geological model.
The use of these data holds out the prospect of the initial geological
models being “closer to reality” than would otherwise be the case.
This work is potentially very time-consuming but there is the hope
that the resulting history-matched models will be more useful.
4.4. Issues Related to Reservoir

Simulation/Up-Scaling
When comparing simulator results with observations it is important
to have an understanding of some of the approximations/limitations
that may be inherent in the use of a simulation model. The comments
in this section assume that a conventional finite difference simulator
is being used.
4.4.1. Issues Related to Grid Size/

Numerical Resolution
Issues related to grid size, numerical resolution and up-scaling will
be relevant to a history match exercise. There will be an unavoidable
lack of accuracy in the way that flow is represented with a relatively
coarse grid model. Single- or multi-phase up-scaling is not able to
fully capture all the details of flow at finer scale lengths than are
captured in the coarse model. It may be better to estimate and accept
this level of inaccuracy rather than to try to match historical data
by changing physical inputs (for instance relative permeability) to
compensate for numerical artifacts.
4.4.2. Issues Related to Representation Rate

Variation With Time
Frequently (for reasons of computational speed or because of how
production data are stored) simulation studies assume that well
production and injection target rates can be set to (for example)
monthly average values. It may also be convenient (computationally)
to “write” simulator output at regular intervals and to use this
for comparing historical and predicted values. This is relevant
to pressure and saturation data that we may wish to compare
with values from, for example, RFT and Thermal Decal Time
(TDT) logs.
What effect does this choice of time intervals have on the
comparison of observed and simulated data? We may expect that
saturations would not change rapidly over short periods. Comparison
of observed and simulated saturations would not be influenced too
much by this choice of time intervals.
This may not be the case for pressures. Pressure communication
may be relatively rapid over inter-well distances. The time variation
(over periods of less than a month) can have a significant impact
on pressures at an off-set well. This may impact how RFT data
are modelled. We may want to represent the time variation of well
rates in more detail prior to the time when RFT data is available.
For similar reasons there would be a preference for comparing
RFT data with pressures at the exact time the data were acquired
rather than from simulator data stored at (for instance) monthly
intervals.
4.4.3. Issues Related to Representation of Well

In-Flow in the Simulator
Reservoir simulators do not generally represent the convergent flow
to wellbores in detail. An approach developed by Peaceman 1978 is
used in most commercial simulators to relate grid block and flowing
bottom hole pressures. This approach was developed assuming 2D
flow, a regular grid and uniform reservoir properties but is used fairly
generally. As long as the simulated well rates are close to the actual
rates this should allow flowing BHP data to be directly compared to
simulated data.
Peaceman (1978) also described how, under the same conditions,
grid block pressure could be related to pressures recorded during
a pressure buildup. This approach can be used to compare grid
block pressures with pressure data acquired during relatively short
buildups. The shut-in and buildups will not be simulated directly.
Static pressures acquired when wells have been shut in for long
periods of time may be compared directly to the calculated wellbore

(or indeed grid block) pressures.
A frequently adopted (if somewhat lazy) approach is to plot
available shut-in pressure against calculated bottom hole and grid
block pressures. Adding averages of grid block pressure from cells
adjacent to the well and production rates can make such plots very
useful.
4.5. Details of a “Conventional” Deterministic

History Match Study
4.5.1. Preparatory Work — Definition of Aims
Clearly the first element of the study should be a definition of the
(business) aims of the study. These aims then need to be translated
into plans (budget, resources, timeline, etc.) for a technical study. It
is important that those involved in the study remain aware of the
business aims and are prepared to recommend stopping the study if
those aims prove to be unrealistic.
4.5.2. Data Review/Well Histories

The initial technical element of the study should involve a review of
the available data. This should be aimed at addressing two issues.
First is the quality of the available data for reservoir characterisation;
in particular are there any “gaps” that data acquisition could help
fill. Second is the need to QC the data (especially the production
and surveillance data). This will allow us to exclude bad data and to
form a view as to the level of accuracy of the remaining data.
Consider shut-in Bottom Hole Pressure (BHP) data as an
example. A review of the data may suggest that some data should
be rejected because of:
• large difference between pressures from different gauges;

• pressure gradient in the wellbore not consistent with fluids in the
wellbore;
• indication of leak in the well;
• flowing pressure reported as a shut-in pressure.
The data review would also give an estimate of the accuracy of the
observed data that is being compared to the simulation data. This
would involve:
• review of gauge accuracy;

• review of the accuracy of correcting data from gauge depth to the
depth of the perforations;
• review of the accuracy of well deviation and wire-line depth
measurements.
This process can be quite time consuming.

One area where particular attention should be paid is the
analysis of cased hole logs. These can provide particularly valuable
information for a history match (estimates of flow profiles from
production logs, estimates of in situ saturation from pulsed neutron
logs (PNL) and through casing saturation logs). The utility of
these logs is, however, dependent on their giving estimates that are
representative of conditions in the reservoir (Fig. 4.5.1).
Here a previously “dry” low pressure interval has a PNL response
that suggests water invasion as a result of cross-flow. Clearly, being
able to match saturations from PNL data should be very useful in a
history match. There is, however, a need to ensure that these data
are representative of conditions in the reservoir and conditions that
the simulation model would reasonably be expected to replicate (it
would be unusual to attempt to simulate explicitly a well being shut
in for a short period before running a PNL).
Well Well
flowing shut-in
Low pressure
zone
High pressure
zone
Figure 4.5.1. Illustration of a common problem with PNL interpretation.

A major element of any data is the preparation of well histories.

Well histories would be expected to include the following:
• a review of drilling/logging the well;

• perforation and completion history of the well complete with all
completion schematics;
• a review of cement quality;
• review of all work overs;
• comments on any mechanical problems encountered during the
history of the well;
• reviews/plots of surveillance and production/injection data.
The data required for a history match would place more emphasis on
historical data (e.g. old completions, history of mechanical problems)
than would be needed for continuing operation.
4.5.3. Preparatory Work — Data Acquisition

Opportunities
As we have seen, lack of uniqueness may be a major concern in a
history matching exercise. Following a review of the available data it
is appropriate to ask if additional data or processing would add value
to the exercise. Clearly there may be significant time constraints in
this process.
Firstly, we may require additional data that would reduce the
uncertainty in model inputs. An example of this could be performing
more core flood work to reduce uncertainty in residual oil saturations.
Secondly, we may try to acquire data with the specific intention of
attempting to match the data in the history matching exercise. An
example of this could be acquiring enough pressure data to map
pressure variation over the field.
4.5.4. Preparatory Work —“Classical” Reservoir

Engineering Calculations/Part Field Models
Classical reservoir engineering calculations and “part field” simula-
tion models are a useful precursor to a history matching exercise.
They can help gain insight into which parameters are likely to have
a significant influence on reservoir performance. They can also give

insight into what limitations a full field simulation model may have
due to numerical resolution.
Material balance calculations may usefully be carried out prior to
history matching. This may allow a preliminary characterisation of
aquifer properties and an assessment of whether the observed average
pressure data are consistent with the assumed reservoir volumes and
compressibility values.
4.5.5. Review of the Simulation Model (and

Geological Model)
In the workflow that was described earlier, a review of the reservoir
simulation model prior to commencing the history matching phase
would be a part of many companies’ QC procedures.
Many of the QC checks are relevant to history matching.
• QC of the basic engineering data to the simulation model will help

for views on the accuracy of such data and the extent to which it
would be acceptable to change these data.
Example of relevance to history matching — Reviewing the pore

space compressibility data would help define an acceptable range
of values that could be used in the history match.
• QC of single- and multi-phase up-scaling should help define limits
on how well the (relatively coarser) full field model will replicate
the performance of finer simulation models. It could also involve
an assessment of the limitations of the geo-model and the extent
to which the up-scaled properties could be reasonably varied.
Examples of relevance to history matching — The extent to which

permeability values are uncertain could be assessed, as well as
the extent to which numerical resolution could influence how
fluid movement is modelled. We would want to avoid having
unreasonable expectations as to how well we could match, for
instance, the water cut of a well that had perforations only a few
grid blocks away from a water zone in the simulation model.
• QC of how the structure is represented and how connectivity is

represented in the simulation model can be important. If fault
geometry/juxtaposition is important in determining flow paths in
the reservoir then it is important to form a view of how uncertain
this is and of the scope for connectivity being distorted in the
up-scaling process.
Example of relevance to history matching — Current workflow

makes it relatively easier to produce many models for the variation
of petrophysical properties (porosity and permeability) than the
structural model. This puts a significant premium on having a
“good” structural model and on ensuring that dynamic data have
been considered when the structural model has been constructed.
• Review of how the wells are represented in the model.
The following may be of particular relevance to history matching.
• Are the wells correctly positioned with respect to faults in the

model?
• Are the intervals that are open to flow correctly positioned with
respect to baffles or barriers to flow?
• How well is the permeability height product seen in pressure
transient analysis in agreement with the appropriate model values?
• Where well performance data indicate a mechanical “skin”, is this
represented in the initial simulation model?
This review can be relatively time-consuming.
4.5.6. Outline of Approach to Matching the Model

The major steps of matching a single model of a reservoir to historical
data could be as follows:
(1) Preliminary simulation runs. The aim of these would be to

ensure that historical well injection and production rates can
be achieved. At this stage some effort may be needed to ensure
that average reservoir pressures are very broadly matched.
(2) A review of:

(a) the extent to which model input parameters could be
reasonably changed;
(b) expectations as to how much influence such changes would
have on the simulation;
(c) the extent to which model outputs should be matched.
(3) Based on the above, carry out a series of sensitivity studies
to better understand the relation between model inputs and
outputs.
(4) Attempt to match the model. This may go through a series of
phases:
(a) early efforts may be on broadly matching pressure;
(b) next we may wish to broadly match fluid movement;
(c) next we may wish to start matching pressure and fluid
movement/production in more detail and to attempt to
match well performance.
The extent to which model inputs are changed in this process
has to be constrained, either quantitatively or qualitatively, by
considerations of “reasonableness”. We also need to form, either
quantitatively or qualitatively, a view on how well the data
are matched. During this process, especially if some features of
reservoir performance prove especially difficult to match, we may
choose to review either the reliability of our measured data or
the assumptions of geological modelling.
(5) Further refining of well performance (especially at the end of the
history) and ensuring the model accounts “smoothly” for the
transition between history and prediction.
(6) At this point it would be good practice to have a QC exercise to
assess the fitness for purpose of the history-matched model.
4.5.7. How Well Should We Aim to Match Data?

In this section, we will address the issue of how well we need to
match data. The issues of what data we need to match will be
addressed later. This will be done with reference to an example
Non-
Initial pressure reservoir
gradient
Good
Sand
Depth
Poor
Sand
Oil
gradient
RFT Pressure
Figure 4.5.2. RFT data and reservoir rock quality.
Initial pressure
gradient observed
Match 1
Depth
Match 2
Oil
gradient
RFT Pressure
Figure 4.5.3. Possible matches to RFT data.
involving matching RFT data. Figure 4.5.2 shows some RFT data
taken from the oil leg of a reservoir. Next to the RFT data there is
an indication of reservoir quality.
There are a number of considerations we need to address when
deciding what a good match to the data could be. These include:
• What is the measurement error of the RFT data? This would

represent a lower limit of how well we may wish to match the
data.
• To what extent would considerations of numerical resolution,
choices of how to represent rate variation, etc. influence the quality
of a match that could be expected? Again this would represent a
lower limit of how well we would wish to match that data.
• Which feature of the pressure variation do we wish to match? If

our main interest is in matching the general decline in the reservoir
pressure then we would wish to match pressures to some fraction
of the difference between initial pressure and current pressure.
If, however, we wish to try to better understand the properties
(possibly the vertical permeability or the lateral extent) of the poor
quality sand in the above figure, then we may be more interested
in matching the pressure “break” over the poor quality sand.
The aims we have in matching the data could influence what would
be considered a better fit to the data. Consider the two potential
“matches” to the RFT data shown in Fig. 4.5.3.
If we were interested in getting a good match to the average
pressure decline, then Match 1 would be preferred. If we wanted a
match that honoured the pressure break, then Match 2 might be
preferable.
4.5.8. Assessing the “Goodness” of a Match

In the above, approaches to making assessments of the “goodness
of fit” were discussed. It is possible to make this assessment
qualitatively — for instance on the basis of visual comparison of
observed and simulated data. (In doing this we may wish to use
error bars to visualise the size of uncertainties in the observed data.)
This approach is feasible and, at least for reservoir with a relatively
small number of wells, is widely used.
For reservoirs with a relatively large numbers of wells, or if
we wish to use any computer-assisted approaches to history match-
ing, it is necessary to quantify how well the data are matched. A
common approach would be to use a weighted Root Mean Square
(RMS) error:

wi (xi (measured) − x i (simulation))2 /σi2 , (4.5.1)
where the wi are weights and the σi represents a target level for
fitting the data. Clearly in using such an approach care needs to be
taken in choosing appropriate wi and σi values.
This only measures how well or badly observed data are matched.
We may also want to account for how realistic or unrealistic the
changed input parameters to the simulation model are. This is
discussed by Schulze-Riegert and Ghedan (2007).
4.5.9. Well Controls During the History Match

The available approaches to controlling production wells during the
history period include:
• control on oil (or gas) rate;

• control on liquid rate;
• control on reservoir volume rate.
If a perfect match is achieved then all of these methods will give the
same result because production of all the phases (and pressure) will
be matched. The choice of control mode should be dictated by the
way this influences our ability to match the model.
There are two arguments in favour of using reservoir volume
rates. The first relates to matching the reservoir pressure. Consider
the material balance equation for a reservoir (or for a fault block
within a reservoir) and for convenience assume there is no free gas.
The material balance equation is:
N pBo + W pBw = N Boi[(Bo − Boi)/Boi

+ (cwSwc + cf )/(1 + Swc)∆P ] + W eBw.
(4.5.2)
Assuming constant oil compressibility this becomes,
N pBo + W pBw = N Boi[co + (cwSwc + cf )/

(1 − Swc)]∆P + W eBw. (4.5.3)
The left-hand side of this equation is the reservoir volume production.

If we get this, the stock tank oil initially in place, fluid compressibil-
ities, and the aquifer influx term (We Bw) correct, then we should
get the correct pressure drop. This would suggest that it should be
easier to match pressure by carrying out a series of simulation runs

where the reservoir volume production is “correct”.
The use of reservoir volume production targets may also make it
easier to match water movement. Consider the case of trying to match
simple 1D flow by adjusting reliable permeability curves. Figure 4.5.4
shows water cut development for three cases:
(1) a “correct case”;
(2) a “modified case” with a slightly different reliable permeability
where reservoir volume production is the same as in the “correct”
case;
(3) as above, but where the oil production rate is the same as in the
“correct” case.
In the modified case water breakthrough is later. For the case where
oil rate control is being used, the reservoir volume rate decreases
causing a further delay in breakthrough time. The difference becomes
greater at later time due to the water cut being underestimated and
the reservoir volume rate too low.
Similar considerations would lead to the use of oil rate control,
exaggerating the difference between the “true” and “modified” cases
if the modified case had earlier water breakthrough.
0.9
0.8
0.7
0.6
Water Cut
0.5 TRUE
Qres
0.4 Qoil
0.3
0.2
0.1
0
0 0.5 1 1.5 2 2.5
Time
Figure 4.5.4. Water cut development with time.

Based on the above, the use of reservoir volume rates to control

production would seem to be highly desirable. The benefits would
be:
(1) to make it easier to match pressure, and also possible to match

pressure without matching the production of individual phases;
(2) to make the process of matching water (or gas) production more
“stable”.
We also need to make decisions about how to control injection wells.

For water injection, due to the small compressibility, it should not
make much difference whether the injection is controlled by reservoir
or surface rate.
What about gas injection? If reservoir volume rates were to be
calculated based on a pressure that was too low, then the rate, in
terms of surface volumes, would be reduced. The converse would
apply if the pressure were too high. This could suggest that it would
be preferable to use surface volume constraints.
4.5.10. Initial Simulation Runs

Initial runs of the simulation model and some work to ensure the
simulation model correctly accounts for perforation histories may be
appropriate prior to any sensitivity studies. This work should have
the following aims:
(1) Review average reservoir pressures. If there is a very large

mismatch between simulated and actual reservoir pressure it
is desirable to attempt to better match history (by adjusting
aquifer influx of fault transmissibility for instance).
(2) Ensure that wells are able to produce at their historic rates.
(3) Review production data to see if this indicates any errors in the
perforation history.
If it is not possible to get a very approximate match to reservoir

pressure then the model may not be an appropriate “base case” for
sensitivity studies. Also, if pressures are too low then it may prejudice
our ability to produce wells at their historical rates; the PI of the
completion multiplied by the maximum possible pressure drop may

not be high enough.
If the model has a broadly correct pressure but some wells are
unable to produce (or inject) at historical rates it would suggest
either of the following:
• the kh products are significantly in error;

• the perforated intervals are incorrect;
• there is flow “behind pipe” that is not captured in the model;
• a well has been stimulated (negative mechanical skin?) and this is
not reflected in the model.
4.5.11. Review of Scope for Changing Model

Input Parameters
Before proceeding with a history match it is desirable to review (and
document) the input parameters that could be varied and the extent
to which it would be acceptable to change them. Clearly the types
of changes will be very case specific. It may be appropriate to give
some examples.
Example 1 — Rock compressibility

Based on a review of experimental (SCAL) data and of uncertainties
about stress regimes we could form a view on how uncertain rock
compressibility data are. This could give us:
• a most likely value (or values);

• a range of values that would be acceptable;
• a larger range that would be considered possible but less likely.
Example 2 — General level of uncertainty in permeability

Based on reviews of the geological model, up-scaling methods, and
pressure transient data we should be in a position to form a view
about the accuracy of the estimated average level of permeability in
the model. This could give us a range of “multiplication factors” that
could be applied, either globally or locally.
Example 3 — Local barriers to vertical flow

Based on geological studies we may recognise that there could be
local barriers to flow. These could, for instance, be shales or cemented
layers. The existence of these at wells may be recognised from well
logs and they may be accounted for to some extent in the geological
model (and through that in the simulation model). We may wish
to form a view on whether introducing (or removing) barriers or
baffles to vertical flow in the model should form a part of the history
matching process.
4.5.12. Sensitivity Study

Prior to carrying out a “manual” history match (also before or as
part of a “computer-assisted” match) we may want to carry out a
sensitivity study. In this study we would look at “inputs” that were
uncertain and evaluate the effect of making “reasonable” changes
to them on the simulation results. Choices could be along the lines
of those in Table 4.5.1; it should be stressed that this is just an
example.
Table 4.5.1. Illustrative inputs on uncertainty/sensitivity.
Input Sensitivity Discussion
Permeability ± 50% Based on comparisons of model

permeability values to estimates
from PBU (Pressure Build Up)
kv/kh × 10 and × 0.1 Based on kv/kh estimates from
of base case PBUs and fine scale models
Krw × 0.5 and × 1.5 Based on analysis of SCAL data
Bo (the oil formation No sensitivity PVT (Pressure, Volume and
volume factor) and Temperature) data give us good
oil viscosity estimates
Fault seal Complete seal to We may have very little
no seal understanding of fault seal
potential
Etc. Etc. Etc.
If models have been produced that have a significant stochastic

input it may be valuable to compare results from different model
realisations at this point.
4.5.13. The History Matching Process

In a “manual” history matching it would be common to adopt the
following approach.
• Concentrate first on pressure, then on fluid movement, then on

details of well performance.
• Concentrate first on matching at a field level (e.g. field average
pressure, field water cut), then in more detail.
• First change inputs globally; later make more local variations.
Clearly the approach to matching the data is iterative. For example
• if we change relative permeability to help match water movement

then this will change fluid mobility and hence the pressure match;
• changing the “skin” term on a well to better match well flowing
BHP could change the proportion of in-flow from different layers;
this could in turn change both the fluid movement and the
pressures.
However, this approach is based on the following observations:
• The time scale for pressure communication over reservoir scale

lengths will be much shorter than the time scale for fluid move-
ment. Thus matching pressure data is likely to be more useful in
constraining large scale aspects of the reservoir model.
• Pressure gradients will determine the general direction of flow.
There would be little point in trying to match the detail of fluid
movement before the pressure was broadly matched.
• Many of the inputs that could be changed to better match fluid
movement (reliable permeabilities, local changes to permeability,
modest changes to kv/kh, etc.) may have little effect on the match
to the pressure.
4.5.14. Matching Pressure

The data that we may wish to match include:
• shut-in bottom hole pressure (SIBHP) data,

• flowing bottom hole pressure (BHP) data,
• RFT data,
• interpreted reservoir (or fault block) pressures,
• pressure maps.
RFT data may be especially useful because of their relatively high

accuracy and their ability to give information of baffles/barriers to
vertical flow.
The inputs we would expect to have a major impact on pressure
would include:
(1) aquifer properties,

(2) volumes in place, fluid and pore space compressibility,
(3) bubble point pressure/variation of bubble point spatially (for
reservoirs where pressure falls below the bubble point),
(4) general level of permeability within flow units,
(5) fault transmissibility,
(6) kv/kh and the transmissibility/lateral extend of barriers or
baffles to lateral flow,
(7) the split in production and injection rates between different flow
zones.
In many cases, it may be appropriate to match estimates of average

reservoir pressure early in the matching process. This could involve
mainly (1)–(3) above points of.
Matching pressure gradients can be very valuable. If we are to
correctly capture the direction of flow in the reservoir then we need
to match pressure gradients. This may be influenced by point (1)
(especially how any aquifer is connected to the model) and points
(4) and (5) of above.
If we are interested in matching pressure variation between flow
units and in better understanding vertical communication, then
we would also be interested in points (6) and (7). For point (7),
matching the split in flow between could involve attempting to

match PLT (Production Logging Tool) data. Thus the process of
matching pressure may involve attempting to match the performance
of individual wells in some detail.
4.5.14.1. Pressure match example

Consider a very simple reservoir model as illustrated in Fig. 4.5.5
below:
Figure 4.5.5. Conceptual reservoir cross-section and RFT data.
In this case there is an isolated reservoir (surrounded by shale —

shown in brown) with oil and water columns and a shale, believed to
be a barrier to vertical flow, within the oil column. The first well is
drilled and is produced from above the shale barrier so as to avoid
water production. A second well is drilled and RFT data acquired.
At the same time as this a SIBHP survey is carried out for the first
well.
Results are compared with a simulation model. This comparison
shows the following:
• simulated pressures are a little too high above the shale barrier;
• there is a decline of pressure below the shale barrier that cannot
be easily explained.
How could we approach history matching our model? We can look at

two aspects of this. Firstly how can we explain the data from below
the shale?
• How accurate are the estimates of initial pressures? And how
accurate is the well #2 RFT data? Can we explain the data by
measurement error/inaccuracy?
• Could there be some flow through the shale? We would need
to review geological views on, for example, small faults allowing
communication across the shale.
• Is there scope for flow from below the shale in well #1? This could
be a result of poor cementing in the well. We may wish to review
any cement bond logs. It may also be possible to run additional
logs, for instance temperature logs, to try to see if there is flow of
fluid “behind pipe”.
Secondly how do we go about matching the pressure data?
If we assume that the pressure decline below the shale is real
and not a result of flow behind pipe then matching the pressure data
could involve changing inputs including the following:
• the volumes in place above and below the shale. This could involve
changes in pore volume (porosity, net to gross or gross rock
volume);
• fluid properties (principally oil and water compressibility values);
• rock (pore space) compressibility;
• the factors influencing flow across the shale (shale permeability,
introducing one or more faults into the model to create sand-to-
sand juxtaposition, etc.).
We would expect to be able to match the observed data in more than
one way — the problem of non-uniqueness. For instance, changes in
volumes in place and in fluid and rock compressibility will have very
similar effects on the average pressure.
In terms of primary production, this non-uniqueness may not
be much of a problem — think about a simple material balance
equation. If we intend to water-flood the reservoir then there is a
clear difference between having larger oil in place and having larger
fluid or rock compressibility. Clearly in this case there is a benefit in

reducing the “level of non-uniqueness” by reducing the uncertainty
associated with compressibility values by having more and/or better
experimental data.
4.5.15. Matching Fluid Movement

The data we may wish to match include:
• the production of oil, gas and water;

• open and cased hole logs that indicate fluid saturations;
• estimates of fluid distributions from seismic data — in some cases
4D seismic may provide very useful data;
• interpretations of fluid movement based on production geol-
ogy studies — for example fluid movement maps and cross-
sections/fence diagrams.
(Clearly we need to be mindful of the differences between measured

data and interpretations. Attempting to match interpreted data,
such as water movement maps, represents a way of building on
the understanding of reservoir mechanisms developed by production
geology studies.)
For simplicity the following discussion and the examples deal
with two-phase oil–water problems. When considering the inputs that
could be changed to better match, we should consider the following
features of oil water displacement:
• the broad pressure gradients within the reservoir — these will

determine the direction of flow in the reservoir;
• the factors influencing local displacement — these would include
relative permeability and capillary pressure models;
• the factors influencing volumetric sweep efficiency — these would
include local permeability heterogeneity, the mobility ratio of the
displacement the stability of the displacement, and the tendency
water to cone or cusp.
Assuming that there exists a broadly adequate match to the pressure,

the inputs that could be changed to attempt to match water
movement would include:

• the degree of permeability heterogeneity;
• permeability anisotropy (kv/kh);
• the relative permeability model (Kr end points, saturation end
points, shapes of curves);
• properties of faults and fractures (local flow barriers or enhance-
ment to flow, connectivity, etc.).
Visualisation of the flow in the reservoir model, for example by
viewing a series of saturation plots over time, can be a useful tool
when determining which of these changes are more likely to be
appropriate. In particular, it can act as a check on whether the “style”
of water movement seen in the simulation model is consistent with
our understanding from other sources.
Other examples of visualisation tools could be comparisons of
saturation maps based on simulation and from surveillance data and
“traffic light” plots indicating (as a function of time) wells where
water production is correctly predicted, over-predicted and under-
predicted.
As noted previously, in attempting to match water production
history we need to give consideration to the numerical resolution of
the simulation model. We can look at two examples.
Example 1 — There is early water breakthrough in the simulator

because there are only two cells between a cell that is below the oil–
water contact (OWC) and a cell that is attached to the well. If we
try to fix the problem by, for instance, reducing the end point water
reliable permeability (Krw), then we risk severely distorting how we
model flow everywhere this change is implemented.
Example 2 — A well cones water “in reality” but this cannot be

captured in the simulation model because the lateral resolution of
simulation grid is not fine enough. It may be possible to make the
simulation model cone water, for instance by reducing permeability in
the vicinity well (thus decreasing pressure near the perforations and
promoting the upward flow of water). This would, however, severely
distort well performance.
In both of these examples, there would be scope for increasing

grid resolution close to the wells. If, however, this is not possible, then
it may be better to live with having poor matches to water production
(provided that we understand the cause) rather than distorting model
inputs in a way that could result in unrealistic predictions.
4.5.16. Matching Water Movement — Example 1

The first example of matching water movement is illustrated in
Fig. 4.5.6.
This also shows a comparison of the simulated and actual water
production at the single production wells. The actual water cut
evolution is shown by the solid line and the simulated evolution is
shown in the dashed curve.
How could the model be changed to better match the observed
production data? Two aspects of the mismatch of simulated data
to history are worthy of note. Firstly, although both simulated and
actual water cut development would indicate some layered behaviour
(the water cut increases in two distinct phases), the simulated water
Water
cut
Time
Figure 4.5.6. Conceptual reservoir cross-section + water cut development.

cut could indicate higher flow in the first layer to produce water.
Secondly the simulation model produces water earlier than is actually
the case.
When we seek to match the water cut we need to think about
the other data that are available, e.g.
• PLT data that could give information of flow in different layers;
• PNL data that could give information of swept zone saturations.
What changes could be made to better match the water cut
development?
The split kh product and ϕh product will influence the style of
water cut development, as will the mobility ratio of the displacement.
The timing of water production will be influenced by volumetric and
local sweep efficiencies.
One approach to modelling that data could be as follows.
• If PLT data are available then attempt to match the split
between layers by changing modifying layer permeability values
(but keeping total kh in line with estimates from PBU data).
• Investigate how reasonable changes to reliable permeability end
points and the shape of reliable permeability curves could influence
the time of water production. The following could give later water
production:
◦ decreasing Sorw (and hence decreasing local sweep efficiency);
◦ decreasing Krw — this would make the displacement less stable
(it would tend to reduce volumetric sweep and local sweep
efficiencies).
• Investigate how reasonable changes to permeability and porosity
could change water production. The following could give later
water production:
◦ reducing permeability heterogeneity within the main flow units;
◦ possibly increasing kv/kh;
◦ increasing porosity.
Based on this work we can try to produce a reasonable match to the
data. There are two potential problems.
Firstly, we may find that it is possible to achieve reasonable

matches in several ways — the problem of non-uniqueness. For
example, decreasing Sorw and decreasing Krw could both improve
the match. One would have the effect of reducing the mobile oil
volume and the other would not. These changes could lead to
significantly different views on reserves.
Secondly, we may find that no set of “reasonable” changes to
model inputs allows us to match the observed data. In this case we
may need to review our assumptions. The following are examples of
what we may do.
• Review the production data. Are we sure of the reliability of the

water cut data for instance?
• Review the mechanical condition of the completion. Is there scope
for water being produced from other intervals (e.g. an underlying
aquifer) because of mechanical problems?
• Review our views on how far input parameters could be varied. For
example if we found it impossible to match the production data
without increasing local sweep would we want to review the SCAL
data and the acceptable range of Sorw values?
• Review our basic geological input. For our present case we could
envisage a barrier to lateral flow down dip of the production well
delaying water production. It would be simple to test whether this
would work in a simulation model. We would then have to review
geophysical data to see if this is plausible and to see if it could be
consistent with PBU data.

This is a variant of Example 1. The aim of this example is to illustrate
the importance of matching pressure and hence the general direction
of flow on matching water movement.
In this case there is the possibility, if the fault is not sealing, of
oil influx from the fault block to the right of the figure (Fig. 4.5.7).
Having flow across the fault in the simulation model would tend to
delay water production in the well.
Sealing
fault?
Figure 4.5.7. Conceptual reservoir cross-section.
If we have an incorrect understanding of pressure communication

and the broad pattern of fluid movement, then there is scope for
making inappropriate changes to match fluid movement. Clearly if
we have more information, for instance PBU data giving information
on the nature of the fault, then we would be better able to produce
an accurate model.

In the previous two sections we have looked at changes that could be
made to match water movement for a model where only a single well
is being considered. What would we do in the case of attempting to
match several wells?
Two broad approaches could be possible. Firstly we could
attempt to achieve the best overall match to history by making
global (or “regional”) changes to selected parameters, as discussed
above. Secondly, we may attempt to make many local changes to
better match individual well performance. Both approaches have
some virtues. The first is simpler and may, arguably, lead to a model
with more predictive capability. The second will provide a better
ability to model short term production prediction.
If the former approach is preferred then it may still be appro-
priate to investigate what sort of local model changes would enable
performance to be better matched to individual wells and to ask
whether these changes are in line without understanding the level of

uncertainty in reservoir description.
If a decision is made to introduce local changes to match indi-
vidual well performance then some changes may be more acceptable
than others. Introducing a barrier to flow that is not seen on seismic
data but which is indicated by PBU analysis may be acceptable.
Introducing local changes to reliable permeabilities or local “pore
volume multipliers” may be less acceptable.
4.5.19. Matching Well Pressures/Detailed

Well Performance
Well in-flow or injection performance needs to be modelled. Also we
may wish to match the details of in-flow or injection profiles, as seen
for example on production logs.
An important element of this work would involve matching
flowing BHP data. There may be particular emphasis on matching
the most recent data. Calibrating the model to these data is
important for using the model for prediction. We would also want
to achieve a reasonable match to these data throughout the history
period. If a well is producing from more than one flow unit with
different pressures then changes to well productivity will change the
split of off-take between the different units. This means that matching
well performance cannot be decoupled from matching pressure and
fluid movement.
The most straightforward changes would involve introducing
“productivity index multipliers” for wells. Ideally these multipliers
would be consistent with our understanding of how well the model
matches kh products and our understanding of skin factors.
In some cases, mainly when the perforated interval is represented
by a small number of cells, well productivity may be poorly
represented in the model when there is multi-phase flow. There
could be segregated flow in reality but diffuse flow, and reduced
total mobility, in the model. In such cases exploring the use of “well
pseudos (possible straight line curves) may be appropriate.
We would also wish to model well and tubing performance prior
to using the model for prediction. If the model is doing a good job
of matching flowing BHP and water cut then this matching should
involve making at most rather minor changes to tubing performance
curves.
4.5.20. The Transition to Prediction

It is generally considered appropriate to attempt to replicate the
latter part of the production history using the well controls that will
be used in a prediction case. This would involve running a simulation
model with the wells controlled by Tubing Head Pressure (THP) and
applying appropriate facility limits. If the simulation does a poor job
then it would not be expected to be useful for prediction. If it does
a good job it may be useful for prediction.
4.6. Automatic and Computer-Assisted History

Matching/Multiple Matches
For many years there has been research into methods for automatic
and computer-assisted history matching. A good review and set of
references is provided in the literature of Schulze-Riegert and Ghedan
(2007). Recently there have been significant advances and computer-
assisted matches with commercially available software increasingly
becoming relatively common.
The availability of automatic or computer-assisted matches,
together with advances in geological modelling, have made producing
multiple matched models a more realistic possibility. A degree of
automation in the matching process has the benefit of making the
process less subjective. It also may reduce the risk of a “manual”
history matching process resulting in multiple models that were too
similar.
4.7. Conclusions
Producing history-matched models may allow us to address impor-
tant questions on reservoir development and management. The
quality and utility of the models depends on how well dynamic
data are accounted for in the modelling process. Much of the effort
associated with accounting for the dynamic data should form a
part of the construction of the geological model. This may make

the process of history matching the dynamic model easier and the
resulting model more useful.
The process of history matching should be performed with a
view to making the resulting model fit for purpose. Giving emphasis
to making only reasonable changes to model inputs rather than
“forcing” outputs to fit observed data may help achieve this aim.
References
Mattax, C. C. and Dalton, R. L. (eds.) (1990). Reservoir Simulation, SPE
Monograph Series, Volume 13, Richardson, Texas.
Peaceman, D. W. (1978). Interpretation of well block pressure in numerical
reservoir simulation, SPE Journal, 18(3), 183–194.
Schulze-Riegert, R. and Ghedan, S. (2007). Modern Techniques for History
Matching, Proceedings of the 9 th International Forum on Reservoir
Simulation, 9–13 December, Abu Dhabi, UAE.
August 11, 2017 10:20 Topics in Reservoir Management - 9in x 6in b2764-index page 245
Index
A Buckley–Leverett, 61, 84, 87

Buckley–Leverett equations, 80
a tie line, 72
Buckley–Leverett solution, 79
absolute permeability, 32
bulk volume, 2, 42–43
actions, 135, 205
bundle-of-parallel-tubes model, 29, 37
adsorbed, 93
bundle-of-tubes model, 30
adsorption, 95
business context, 211
analogue models, 110
anisotropy, 10
C
aquifer treatment, 197
Archie, 44 capillarity, 18, 55, 57
Archie’s law, 39–40 capillary number, 55, 58
Archie’s second law, 40 capillary pressure, 21–22, 27, 29–30,
Arctic, 101 34, 37, 46, 111, 133–134, 148,
arithmetic mean, 15 153–156, 161, 174–179
ASP, 92 capillary rise, 25–26
capillary trapping, 67
automatic history matching, 211
capillary tube, 25
capillary-gravity equilibrium, 27
B
carbon dioxide (CO2 ), 50, 67, 79, 102
barrels, 45 carbonates, 39
Bear, 14, 44–45 Carter Tracy, 199
Beran, 18, 44 Cartesian system, 138–139, 141, 144
Bernoulli’s equation, 11 cement bond logs, 235
black oil fluid properties, 190 cementation index, 39
bottom hole pressure, 219 centipoise, 12
brine, 37–38, 41 CH4 , 67
bubble, 21 chalk, 12
bubble point curve, 70 channelling, 60, 65, 93
Buckley–Leverett displacement, 131, characteristic velocity, 99
153, 172 chemical, 49
245
chemical EOR, 50 desalination plants, 102

chemical equilibrium, 68 developing miscibility, 73
chemical techniques, 57 dew point curve, 70
circular pore tubes, 12 diatomites, 4
clay, 12, 40 differential liberation, 186, 188–190,
CO2 injection, 52 193, 196
CO2 sequestration, 64 diffusivity equation, 117–118,
comprehensive models, 110–111 137–138, 199
compressibility, 42–43, 117, 122, 148, dimensionless numbers, 65
150–151, 186, 188, 193, 196 dispersion, 65
computer-assisted history matching, dispersion and weighting, 125
243 divalent cations, 97
condensing gas drive, 76–78 drainage, 33
conductivity(ies), 36–37, 41 dual-porosity, 3
confining pressure, 42–43 Dullien, 13, 44
connate water bank, 98–99 Dvorkin, 44
connected volume factor, 55 Dykstra and Parsons, 65
constant composition expansion, dynamic data, 216
186–187, 189–190, 194
constraints, 125, 135–137, 142, 146, E
203–206 economic efficiency factor, 55
contact angle, 22–23, 46 effective, 17–18
core data, 148, 159 effective permeability, 14–15, 45
corner point representation, 144 effective porosity, 4
crack, 44 electrical double layer, 41
critical tie line, 78 electrical resistivity, 2, 35–36
cross-sectional models, 136, 161, elemental models, 110–112
169 empirical model, 61
cylindrical single well models, 137 end-point relative permeabilities,
cylindrical systems, 140, 142 34
engineering, 2, 10, 34
D enthalpy, 11
Daqing field, 94 entropy, 19–20, 22
Darcy unit, 11 EOR, 49
Darcy’s law, 7–11, 13, 16, 32, 35, 37,
45, 112, 115–116, 121–122, 149, F
151, 198 fence diagrams, 216
Darot, 44 Fetkovitch, 199–200
data acquisition opportunities, 221 field scale deployment, 94
data manipulation, 161, 193 finite difference, 114, 118–119, 125,
de Marsily, 44 127, 134, 137, 150, 174, 178–179,
define EOR, 48 190, 197
definitions, 183 first contact miscible, 73, 77
density, 41, 45–46 first contact miscible WAG, 85
depositional environment, 65 fluid compressibility, 41
Index 247
fluid potential, 10 heavy, 49

Fontainebleau, 39 heterogeneity, 4, 6, 17–18, 34, 59–60,
Fontainebleau sandstones, 40 65
Forchheimer, 11 heterogeneity index, 66
formation, 40 heterogeneous, 5–6
formation factor, 39 high sulphate, 49
formation resistivity factor, 38 history match, 205–207, 209
fractional flow, 85, 87, 91, 94, 98 homogeneous, 4, 16, 46
fractional flow curve, 86, 99 homogeneous rock, 15
fractional flow function, 81 horizontal, 9, 14, 16, 66
fractional flow theory, 81 horizontal flow, 10
fractured, 3 horizontally, 15
fractured reservoirs, 34 Hurst Van Everdingen, 198
fractures, 3 hydrocarbon, 50, 79
free, 28 hydrocarbon gas, 101
free water level, 28 hydrocarbon types, 182–183
full field models, 137, 170
I
G
imbibition, 33–35
gas, 1, 7, 11, 18–19, 21–22, 24–25,
immiscibility, 71
32, 42, 50, 100
immiscible, 77
gas injection, 49, 51, 97
immiscible displacement, 79
gaseous, 23
immiscible fluid, 80
geocellular up-scaling, 150
implicit and explicit formulation, 121
geographic, 51
improved oil recovery (IOR), 48
geographic deployment, 49
ineffective porosity, 4
geological heterogeneity, 64–65,
infill, 66
103
geometric mean, 17 initialisation, 134, 175–176, 179
global oil production, 50 injectivity, 94
gneiss, 12 instability, 60, 66
goodness of a match, 226 interface, 18–24
grains, 5, 9, 40 interfacial, 24
granite, 12 interfacial tension, 58
gravel, 12 interference test data, 216
gravity, 26 internal, 18–19
gravity override, 66 internal energy, 19–21
gravity segregation, 63 inverse problems, 210
gravity tongue, 64 irreducible, 30, 46
grid block pressure, 218 irreducible saturations, 34
Gueguen, 44 irreducible water saturation, 33
isotropic, 13
H
harmonic, 18 J
harmonic mean, 17 J-function, 31, 46
K mixed wet, 57
Kozeny–Carman equation, 14 mobility, 37
mobility control, 66
L mobility ratio, 61, 93
model capillary pressure, 174
layered, 15, 17
model components, 134
layered reservoir, 45
model control, 134
layered rock, 14, 16
model fluid properties, 190
Leverett, 30, 44
model relative permeability, 161
light tight oil, 52
model well and production data, 201
limestones, 12
Mukerji, 44
linear solvers, 126–127, 130
multi-contact miscibility, 75
liquid, 19
multi-contact miscible gas, 103
local grid refinement, 144–145, 147
multi-contact miscible processes, 74
log data, 148, 157
multilateral, 66
logging, 36
multiple models, 211
Lorenz coefficient, 66
low salinity, 49, 103
N
low salinity water, 50, 101
low salinity water flooding, 92, 97 non-orthogonality, 147
non-wetting fluids, 56
M non-wetting liquid, 24
non-wetting phase, 33
macroscopic sweep, 63, 65–66, 103
nonlinearity and outer iterations,
macroscopic sweep efficiency, 59
126
Magnus, 102
numerical aquifer, 200
manipulation of capillary pressure,
numerical models, 112
176
numerical resolution, 217
mass conservation, 113–114, 117
matching fluid movement, 236
O
matching pressure, 233
matching water movement, 238 Ohm’s law, 36–38
material balance calculations, 42–43 oil, 1–4, 6–7, 21–22, 25–26, 28, 30,
mathematical models, 110, 112 32–36, 40–41, 45–46
Mavko, 44 oil supply, 53
method of characteristics, 81, 84 oil wetting, 57
microscopic displacement, 67 oil–water contact, 26, 28
microscopic displacement efficiency, oil–water transition zone, 27
59, 94, 97, 103 oil-wet, 25
microscopic displacement factor, 58 oil-wet relative permeabilities, 99
microscopic fluid distribution, 56 oil-wet rocks, 40
mineral grain volume, 42 optimum WAG ratio, 90
mineral volume, 2 output, 115, 135
minimum miscibility pressure, 101
miscible, 103 P
miscible displacements, 60 parallel-tube model, 27, 34, 39, 45–46
miscible gas, 101 part field models, 221
Index 249
permeability(ies), 1–2, 7–8, 10–14, R

17–18, 32, 37, 39, 45, 111, 133, 136, rarefaction, 98
148–153, 155, 157, 159–165, recovery efficiencies, 55
167–174, 176–179, 202 recovery factor, 53, 55, 59
permeability tensor, 10 relative, 33
phase contact, 68 relative permeability, 32, 58, 111, 148,
phase equilibrium, 68 151–153, 155, 157, 161–163, 165,
physical models, 110–111 168–174, 176–179, 202
Plait point, 73, 78–79 relative permeability end-points, 34
Poiseuille’s equation, 13, 35 representative elementary volume,
polar components, 97 4–5
polyacrylamide, 94, 101 representative volume element, 5
polymer, 49, 92, 97, 101–102 reservoir(s), 10–12, 14, 17, 25, 27,
polymer flooding, 93–94 30–31, 33, 35, 40, 42, 45–46
polymer gels, 66 reservoir description, 112–114, 134,
pore compressibility, 2, 42–44 137, 205
pore doublet, 34 reservoir engineering, 1, 36
pore pressure, 2, 43–44 reservoir models, 109, 113–114, 134,
pore scale displacement, 55 147, 168–169, 174–175, 195, 197,
pore size, 5, 12–14, 39, 55 202, 205
pore space, 42 reservoir simulation, 6
pore volume, 2, 42–43 reservoir sweep, 55
pore-size distribution, 29 residual non-wetting phase, 58
porosity, 1–6, 9, 12–13, 39–40, 45, residual oil saturation, 33, 35
117, 126, 133, 148–151, 157–158, residual saturation, 34, 56–57
161–162, 177–178 resistivity, 36–37, 40
potential, 16 retardation factor, 95
practical considerations, 205 REV, 6
prediction, 113–114, 205–207 Reynolds number, 45
pressure(s), 2, 7–8, 12, 19, 21, 26, 32, routine core analysis, 148
41–42, 45 Ruffet, 39–40, 44
pressure gradient, 9
pressure transient data, 216 S
pressure-composition diagram, 69 salts, 37
primary drainage, 34 sand grains, 7
primary porosity, 3 sandstone, 3, 5–6, 12, 39, 44
production control data, 203 saturation, 1–2, 6, 29–30, 32, 45
production data, 114, 135, 201, saturation exponent, 40
203 Scheidegger, 14, 44
properties of mixtures, 180 screening criteria, 49
Prudhoe Bay, 67, 101 secondary porosity, 3
pseudo component, 68, 86 sector models, 136
pseudo relative permeability, 161, seismic compressional waves, 43
168–171, 173 sensitivity study(ies), 224, 231
pulsed neutron logs, 220 separator tests, 190, 194
shale oil, 52, 100 tortuosity, 14, 39

shales, 4, 12 total compressibility, 43
shaly, 41 total porosity, 3–4
shaly sands, 40 transition, 29, 46
shear thinning, 93 two-component system, 84
shock, 82, 98 two-phase envelope, 72
shock front, 87–88, 96, 99 two-phase flow, 32
shock front mobility ratio, 60 two-phase multi-component
shock velocities, 84 displacements, 67
shutting off, 66
silt, 12 U
single component properties, 180 Ula, 101
sketch, 19 uncertainty, 65, 103
snap off, 56 uncertainty assessment, 211
solvent flooding, 58 unconventional oil, 100
spatial variations, 134, 196–197 unconventional production, 49
special connections, 142, 144, 146 unstructured grids, 145–146
special core analysis, 148, 150, 152 up-scaling, 217
specific surface, 14
stable velocity, 62 V
static and dynamic data, 215 vapourising, 78
static model, 216 vapourising gas drive, 76–77
steam injection, 49–51 variable Api gravity, 197
stream line grids, 141 variable bubble point, 197
subsidence, 43 vertical, 16
surface constraints, 146 vertical equilibrium, 166, 168, 173,
surface tension, 2, 18–21, 23, 26, 46 178–179
surfactants, 49, 101 vertical flow, 9
vertical permeability, 10
T vertical segregation, 59
targets, 203 viscosity, 7–8, 11, 32, 37, 45, 58
temperature, 19, 41 viscosity ratio, 60, 62, 65
tension, 22, 46 viscous, 66
ternary diagram, 68, 78 viscous cross-flow, 93
test data, 118, 148, 160 viscous fingering, 65, 93
thermal, 49 viscous fingers, 60–61
thermal methods, 51 viscous instability(ies), 59, 64–65
thermodynamic equilibrium, 72 Vosges, 39–40
three-phase flow, 90–91
three-phase relative permeability, 92, W
165 water, 1–2, 6–7, 21–22, 25–28, 30,
tie line, 84 32–35, 37, 40, 45–46
tight oil, 53 water alternating gas (see also
tight reservoirs, 49 WAG), 63–64, 66, 103
to extra heavy, 49 water-wet, 24–25, 29–30, 33, 40, 56
Index 251
water-wet relative permeabilities, 99 X

water-wet reservoir, 28 xanthan, 101
Waxman–Smits equation, 41 xanthan gum, 94
well controls, 227
well histories, 219 Y
well inflow, 201
Young–Laplace equation, 21–22, 27,
well pseudo relative permeability, 173
34
well workovers, 102
well-test analysis, 43
Z
wettability, 2, 18, 57
wetting fluids, 56 Zimmerman, 44
wetting liquid, 24
wetting phase, 33

The Imperial CollegeLecturesinPetroleumEngineeringVolume 3

Uploaded by

Copyright:

Available Formats

The Imperial CollegeLecturesinPetroleumEngineeringVolume 3

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

The Imperial CollegeLecturesinPetroleumEngineeringVolume 3

Uploaded by

Copyright:

Available Formats

What topics are covered in the book?

What topics are covered in the book?

What are some key concepts discussed?

What are some key concepts discussed?

The Imperial College Lectures in

Q0082_9781786342843_TP.indd 1 11/8/17 11:33 AM

This page intentionally left blank

b2530_FM.indd 6 01-Sep-16 11:03:06 AM

Q0082_9781786342843_TP.indd 2 11/8/17 11:33 AM

Library of Congress Cataloging-in-Publication Data

British Library Cataloguing-in-Publication Data

Copyright © 2018 by World Scientific Publishing Europe Ltd.

Desk Editors: Dipasri Sardar/Mary Simpson/Shi Ying Koe/Jennifer Brough

Typeset by Stallion Press

Dipa - The Imperial College Lectures - V3.indd 1 26-07-17 3:51:51 PM

vi Topics in Reservoir Management

mechanical engineering, and little if any prior exposure to petroleum

About the Authors

Samuel C. Krevor is a Senior Lecturer in the Department of

Ann H. Muggeridge holds the Chair in Reservoir Physics and EOR

viii Topics in Reservoir Management

indicator of reservoir compartmentalization. She has a BSc (Hons) in

David Waldren is director of Petroleum Consulting and Training

Robert W. Zimmerman obtained a BS and MS in mechanical

Chapter 1. Introduction to Rock Properties 1

x Topics in Reservoir Management

1.4.4 Capillary Rise . . . . . . . . . . . . . . . . . . . . . . 25

Chapter 2. Introduction to Enhanced Recovery

2.1 Introduction: Deﬁnition, Techniques and the Global

2.3.2.1Review of the Buckley–Leverett

Chapter 3. Numerical Simulation 109

3.1 Reservoir Models . . . . . . . . . . . . . . . . . . . . . . . . . 109

xii Topics in Reservoir Management

3.2.5 Implicit and Explicit Formulation . . . . . . . . . 121

3.7.1 Data Manipulation . . . . . . . . . . . . . . . . . . 161

xiv Topics in Reservoir Management

3.12.2.2 Constraints . . . . . . . . . . . . . . . . . 204

Chapter 4. History Matching 209

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209

Introduction to Rock Properties

2 Topics in Reservoir Management

described and quantiﬁed in terms of properties known as wettability

1.2. Porosity and Saturation

Introduction to Rock Properties 3

The fraction of the cylinder that is occupied by pore space

The porosity of a reservoir rock can range from a few percent, to

4 Topics in Reservoir Management

1.2.2. Heterogeneity and “Representative

Figure 1.2.2. Sand–shale sequence in a reservoir. Such a reservoir is homogeneous

Introduction to Rock Properties 5

Figure 1.2.3. (a) Porosity at x1 is 0, and at x2 is 1, illustrating that porosity

At the other extreme of length-scale, all porous rocks are

6 Topics in Reservoir Management

the porosity, we are usually implicitly referring to the porosity as

Sw = Vw /Vp , So = Vo /Vp . (1.2.3)

The saturation of each phase must, by deﬁnition, lie between 0 and 1.

Introduction to Rock Properties 7

the case that,

If there is also some hydrocarbon gas in the pore space, then

1.3. Permeability and Darcy’s Law