See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/318766675

Multi-Tank Material Balance Analysis in Heterogeneous Oil Reservoirs

Conference Paper · July 2017

DOI: 10.2118/189092-MS

CITATIONS READS
0 151

2 authors:

Victor Chinemelu Molokwu M.O. Onyekonwu

6 PUBLICATIONS   7 CITATIONS   
University of Port Harcourt
131 PUBLICATIONS   456 CITATIONS   
SEE PROFILE
SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Gas Hydrate Inhibition using Local Inhibitors in a Simulated Offshore Environment View project

Research project View project

All content following this page was uploaded by M.O. Onyekonwu on 15 June 2020.

The user has requested enhancement of the downloaded file.

SPE-189092-MS

Multi-Tank Material Balance Analysis in Heterogeneous Oil Reservoirs

V. C. Molokwu, Laser Engineering and Resources Consultants Limited; M. O. Onyekonwu, University of Port
Harcourt

Copyright 2017, Society of Petroleum Engineers

This paper was prepared for presentation at the Nigeria Annual International Conference and Exhibition held in Lagos, Nigeria, 31 July – 2 August 2017.

This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents
of the paper have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect
any position of the Society of Petroleum Engineers, its officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written
consent of the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words; illustrations may
not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.

Abstract
This paper presents detailed theories and method of determining fault transmissibility and production
history matching using the multi-tank material balance. This approach uses a two-step optimization process
whose algorithm can be written as macros in spreadsheets. The upper-level (outer) stage optimizes the
transmissibility and hydrocarbon initially-in-place while the lower-level (inner stage) optimizes the pressure
of the support tank at each time step. This approach is validated using numerical simulation. This approach
will be highly beneficial in effective reservoir management where little or no 3D seismic exists and for
cases of sparse production data.

Introduction
Material balance has since been recognized as one the methods used by petroleum engineers for interpreting
and predicting reservoir performance. The material balance is a zero dimensional model; it considers the
reservoir as a tank having the same pressure and PVT properties throughout, at a particular time. However,
there could be substantial pressure variations across the reservoir and if these are not accounted for, could
lead to rigorous history matching of a highly reliable production data. Pressure variations across a reservoir
are a result of compartmentalization of the reservoirs – where two or more reservoirs are in hydraulic
connection with one tank providing energy support to the other tanks. A compartmented reservoir could be
a faulted reservoir made up of two or more fault blocks and connected by partially sealing faults (Hagoort
and Hoogstra 1999); it also include reservoirs that have distinct geological units separated by intra-reservoir
shales (Mogbolu et al. 2015).
Compartmented reservoir can be effectively modelled using multi tanks which are connected to one
another by a permeable barrier. Each tank has its own material balance and common to both tanks is the
reservoir production transfer across the permeable boundaries. A major determinant of this boundary rates
is the transmissibility across the tanks which is a characteristic of the fault.
A fast and reliable estimate of reserves in compartmented reservoirs is also possible using multiple tanks
existing at its own pressure equilibrium. This serves as a guide for future development actions and to obtain
reliable estimates – hydrocarbon initially in place and fault transmissibility require history matching of
production data from all compartments. There exist numerous literatures on material balance equation for
2 SPE-189092-MS

compartmented gas reservoirs but few for oil reservoirs. Compartmentalization of gas reservoirs is a solution
to history matching tight gas reservoirs where there is always difficulty in obtaining average reservoir
pressure. Payne (1996) applied the multicompartment reservoir model to single, tight gas reservoirs. Payne's
method is a simple and straight forward method and ignores the changes in the flow across the boundaries
and gas properties during a timestep. Haggort and Hoogstra (1999) presented another simple but rigorous
numerical method for the solution of the material balance equations for compartmented gas reservoirs. Their
method employs an implicit calculation scheme for calculation of the pressures of individual tanks during
a timestep.
In the same vein, the compartmented approach is applied to oil reservoirs and solved using a nested
optimization process whose algorithm can be written as macros in spreadsheets. This is an iterative process
and consists of an upper-level and a lower-level optimization tasks. The upper-level (outer) stage optimizes
the hydrocarbon-in-place, transmissibility and also aquifer parameters while the lower-level (inner stage)
optimizes the pressure of the various compartments for each timestep.
The aim of this paper is to present theories involved in transmissibility determination using a
multi-tank material balance. It also aims at applying an unconstrained optimization technique to
transmissibility determination and production history matching using synthetic production data. This
technique provides reservoir engineers and software developers with a method that determines the extent
of compartmentalization especially in sparse production data and when no seismic data is readily available.

Theories of Multi-Tank Material Balance

Application of the multi-tank material balance to compartmentalized reservoirs can determine the
transmissibility of the fault and improves the initial estimates of the hydrocarbon in place in the main tank.
This paper considers a two-tank system as shown in Fig. 1. Each tank exists at its own pressure equilibrium
and connected by a semi-permeable barrier.

Figure 1—A two-tank system

The material balance equation for both tanks assuming tank 2 provides energy (pressure support) to tank
1, is given by
For tank 1:

(1)

For tank 2:

(2)

The reservoir rate between tanks is given by

SPE-189092-MS 3

(3)

The rate is zero at initial condition and as such, both tanks exist at equal potentials. Any difference in
pressures between tanks is purely due to hydrostatic effects. Such difference is as a result of the tanks being
at different depths (Fig. 2).
(4)

The rate can also be zero when both tanks exist at different potentials at initial condition. This is due to
an actual potential difference supported by the pressure threshold of the fault (Fig. 2). In such case,
(5)
Φ2 – Φ1 ≥ Pth for fluid to flow

Figure 2—Possible initial conditions.

The cumulative influx between tanks is expressed by

(6)

is the average mobility of a fluid phase.

Np21 can be expressed using a transmissibility modifier as

(7)

C21 is the transmissibility between compartments and is due to mobility contrast between them. It assumes
the interface to be stationary with no thickness; change of reservoir properties is abrupt and there is no
resistance to flow between the compartments. It is the harmonic average of each tank's transmissibility.
The transmissibility of a tank, expressed in field units is given by

(8)
4 SPE-189092-MS

(9)

The average mobility of the jth fluid phase is given by

(10)

The total mobility is given by

It is usually recommended to start with a single tank model. Hydraulic communication between reservoirs
can be inferred from the Campbell's diagnostic Plot. An initial flat line in data trend which later on
increases, signifies additional pressure support from a different source. This increasing trend does not start
immediately with production unlike an aquifer support which starts immediately. The delay in increase
for a compartmented reservoir is due to the presence of a semi-permeable barrier (fault plane) which has
significant thickness. The threshold of the fault plane controls the time of communication between both
tanks; no barrier exists in a reservoir-aquifer system. A reservoir-aquifer system could also be modelled
using the two-tank model; in such case, Γ ≅ 1

The Solution Technique

A fast, robust, two-step optimization technique is used for solving the systems of equations. The upper-
level cost function is minimized by use of the Levenberg-Marquardt algorithm (LMA). The LMA is an
iterative process that locates the minimum of a multivariate function that is expressed as a sum of squares
of a non-linear real valued function (Lourakis et al. 2005). LMA is a blend of the steepest descent and the
Gauss-Newton minimization processes. When the current solution is far from an optimum, it behaves like
a steepest descent method. When the current solution is near optimum, it behaves like a Gauss-Newton
method. At the upper level, the variables (parameters) to be optimized can be the hydrocarbon initially-in-
place in the compartments, the transmissibilities and also aquifer parameters.
The lower-level is a Newton-Raphson minimization process. The parameters to be optimized are the
pressures of the adjoining compartment. The Newton-Raphson algorithm is adopted at the inner stage as
it is a single-parameter estimation.

Application to the Two-Tank Model

This paper considers tank 1 as the main tank. An undersaturated condition, with no water influx is assumed
to exist in both tanks throughout the given production period. Thus, Eqs. 1 and 2 become
(11)

(12)
Combining Eqs. 11 and 12 gives
(13)
The average mobility due to the jth fluid phase becomes

(14)
SPE-189092-MS 5

Since only one phase exist in both tanks at all conditions, total mobility,
Further details are shown in the Appendix.

Figure 3—Flow chart of the optimization process.

Validation of the Method

A 10 × 5 × 10 grid simulated model with two equilibrium regions was built to simulate the cumulative
production and tank pressures for the two-tank model. The transmissibility between the two equilibrium
regions was set with a transmissibility modifier of 0.3 and a pressure threshold of 800 psia. The simulation
grid is shown in Fig. 4 and grid properties used in simulation are shown in Table 2. Two wells, P1 and P2
are placed in regions 1 and 2 respectively. A porosity value of 0.5 is used in P2 block. Well P2 is shutin to
ensure the region acts as a pressure support to region 1.
6 SPE-189092-MS

Figure 4—10 × 5 × 10 grid blocks with two equilibrium regions

PVT properties (i.e. formation volume factor and viscosity) of the fluid are estimated using Vasquez-
Beggs and Petrosky-Farshad correlations respectively. Unknown data points in the PVT checklist (Table 2)
are obtained using the Akima cubic spline.

Table 1—Simulation grid properties

SPE-189092-MS 7

Table 2—PVT checklist

Table 3—Reservoir production data

Hydraulic communication between reservoirs is inferred from the Campbell's diagnostic plot in Fig. 5.
From Fig. 5, the trend with cumulative production is initially flat and later deviates as production proceeds.
This indicates that a barrier has been broken between tanks and one tank is acting as a pressure support
to the other tank. The production data obtained from simulation is thus modelled by use of the multi-tank
material balance. The results are shown in Table 4.
8 SPE-189092-MS

Figure 5—Campbell's Diagnostic Plot

Table 4—Results obtained from optimization

From Table 4, it can be seen that the best parameters for the oil initially-in-place and transmissibility are
very close to values which was set in simulation. Also, the estimated pressures in tank 2 matches closely
the values obtained from simulation. The match of the pressure with time is shown in Fig. 6.
SPE-189092-MS 9

Figure 6—Tank 2 pressure match

C12 is estimated as using Eqs. 8 and 9,

Conclusions
1. A statistical method of determining fault transmissibility and production history matching in
undersaturated oil reservoirs using the multi-tank material balance is presented.
2. This optimization technique provides reliable estimate of the fault transmissibility and improves the
hydrocarbon-in-place in the main tank. It also determines the pressure changes in the support tank.

Acknowledgements
The authors appreciate the suggestions and discussions of the research and development team of Laser
Engineering during the preparation of this manuscript.

Nomenclature

Dimensionless and Field Variables.

A = Cross-Sectional Area of a Tank, ft2
Bo = Oil Formation Volume Factor at Average Reservoir Pressure, RB/STB
Bw = Average Water Formation Volume Factor, RB/STB
cf = Formation Compressibility, psi−1
cw = Formation Water Compressibility, psi−1
C12 = Transmissibility between Compartments due to Mobility Contrast, Rb/D·cp/psia
Eo = Oil Expansion Factor, RB/STB
10 SPE-189092-MS

Efw = Rock/Fluid Expansion Factor, RB/STB

GOR = Instantaneous Gas-Oil Ratio, scf/STB
h = Net Pay Thickness, ft
k = Formation Permeability, md
kr = Relative Permeability to a Fluid Phase
L = Length of a Tank, ft
N = Original-Oil-in-Place, STB
Np = Cumulative Oil Produced, STB
Np21 = Cumulative Fluid Influx from Tank 2 to Tank 1, STB
p = Average Reservoir Pressure, psia
Pth = Threshold Pressure, psia
qo = Oil Production Rate, STB/D
q21 = Reservoir Fluid Rate from Tank 2 to Tank 1, RB/D
Rp = Cumulative Gas-Oil Ratio, scf/STB
Rs = Gas Solubility, scf/STB
Swi = Initial Water Saturation, fraction
t = Production Time, days
Δt = Production Time Step, days
T = Transmissibility of the Fault, Rb/D·cp/psia
Wp = Cumulative Water Produced, STB
Δz = Depth Difference between Tanks, ft

Greek Variables
= Average Fluid Gravity in both Tanks, SG
Γ = Fault Transmissibility Multiplier
λd = Levenberg-Marquardt Damping Factor
= Average Mobility of a Fluid Phase
= Total Mobility of the Fluid
μ = Fluid Viscosity at Average Reservoir Pressure, cp
ϕ = Formation Porosity, fraction
Φ = Fluid Potential, psia

Subscripts
1 = Tank 1
2 = Tank 2
i = Initial Conditions
j = Production Data Point Index; Fluid Phase Index
m = Number of Production Data Points of the Main Tank Excluding the Initial Conditions
n = Number of Pressure Support Tanks (Includes the Main Tank)

Subscripts
k = Loop Control Variable

References
1. Akima, H. 1969. A Method of Smooth Curve Fitting. ESSA Technical Report. Institute for
Telecommunication Sciences, Boulder, Colorado. ERL 101-ITS 73
SPE-189092-MS 11

2. Anderson, D. and Mattar, L. 2004. Practical Diagnostics Using Production Data and Flowing
Pressures. Presented at the Annual SPE Technical Conference and Exhibition, Houston, Texas,
26-29 September. SPE-89939-MS. http://dx.doi.org/10.2118/89939-MS.
3. Draper, N. R. and Smith, H. 1981. Applied Regression Analysis, Second Edition. New York: John
Wiley and Sons.
4. Havlena, D and Odeh, A. S. 1963. The Material Balance as an Equation of a Straight Line. J Pet
Technol 15 (08): 896–900. SPE-559-PA. http://dx.doi.org/10.2118/559-PA.
5. Havlena, D and Odeh, A. S. 1964. The Material Balance as an Equation of a Straight Line—Part
II, Field Cases. J Pet Technol 16 (07): 815–822. SPE-869-PA. http://dx.doi.org/10.2118/869-PA.
6. Hagoort, J and Hoogstra, R. 1999. Numerical Solution of the Material Balance Equations of
Compartmented Gas Reservoirs. SPE Reservoir Eval. And Eng. 2 (04) 385–392. SPE-57655-PA.
http://dx.doi.org/10.2118/57655-PA.
7. Lourakis, M. I. and Argyros, A. A. 2009. A Software Package for Generic Sparse
Bundle Adjustment. ACM Transactions on Mathematical Software. 36 (01) 1–30. http://
doi.acm.org/10.1145/1486525.1486527.
8. Marquardt, D. W. 1963. An Algorithm for Least-Squares Estimation of Nonlinear Parameters.
Journal of the Society of Industrial and Applied Mathematics 11 (02): 431–441.
9. Mogbolu, E., Okereke, O., Okporiri, C. et al. 2015. Using Material Balance (MBAL) Multi-
Tank Model to Evaluate Future Well Performance in Reservoirs with Distinct Geological Units.
Presented at the Nigeria Annual International Conference and Exhibition, Lagos, 4 – 6 August.
SPE-178484-MS. http://dx.doi.org/10.2118/178484-MS.
10. Ojo, K. P. 2006. Material Balance Revisited. Presented at the Nigeria Annual International
Conference and Exhibition, Abuja, 31 July – 2 August. SPE-105982-MS. http://
dx.doi.org/10.2118/105982-MS.
11. Ojo, K. P., Tiab, D. and Osisanya, S. O. 2006. Dynamic Material Balance Equation and Solution
Technique Using Production and PVT Data. J Can Pet Technol. 45 (03): 45–53. PETSOC-06-03–
03. http://dx.doi.org/10.2118/06-03–03.
12. Payne, D. A. 1996. Material Balance Calculation in Tight Gas Reservoirs: The Pitfalls of
p/z Plots and a More Accurate Technique. SPE Reservoir Engineering. 23 (02): 209–218.
SPE-10356-PA. http://dx.doi.org/10.2118/10356-PA.
13. Transtrum, M. K., Machta, B. B. and Sethna, J. P. 2011. Geometry of Nonlinear Least Squares
with Applications to Sloppy Models and Optimization. American Physical Society. 83 036701:
1–35. http://dx.doi.org/10.1103/PhysRevE.83.036701.
14. Ukwu, A. K. and Onyekonwu, M. O. 2014. Advancement in Material Balance Analysis.
Presented at the Nigeria Annual International Conference and Exhibition, Lagos, 5 – 7 August.
SPE-172415-MS. http://dx.doi.org/10.2118/172415-MS.
12 SPE-189092-MS

Appendix A
Definition of Terms
Further terms used in this paper as defined as follows
The underground withdrawal is given by
(A-1)
The oil expansion factor is given by
(A-2)
The rock and connate water expansion factor is given by

(A-3)

The derivative of the oil expansion term is given by

(A-4)

The derivative of the rock and connate water expansion term is given by

(A-5)

The vector of parameters

(A-6)
SPE-189092-MS 13

Appendix B
The Upper-Level Optimization Process
The total cost function of the main compartment at the upper-level optimization is given by

(B-1)

Where
(B-2)
θ is a p×1 vector of parameters (i.e., hydrocarbon initially-in-place, transmissibility etc.)
Pj is a n×1 vector of pressures acting on the main tank at the jth period.
The metric of the LMA is given by
(B-3)
Where J is the Jacobian matrix; λd is the damping factor.
The cost gradient is given by
(B-4)
The step in the LMA algorithm is given by
(B-5)
The current vector of parameters,
k is the iteration steps. k = 0, 1, 2, 3, …, ns
For each level, the iteration continues until a set of conditions are met.
Considering the main tank equation given by
(B-6)
Np21 is evaluated by use of the trapezoidal rule.
14 SPE-189092-MS

Appendix C
The Lower-Level Optimization Process
The lower-level optimization involves solving Eq. C-1 implicitly at each time step to obtain tank pressure, p2.
(C-1)
Eq. C-1 is solved to obtain tank pressures, p2,l,…, p2,m using the system described in Eq. C-2

(C-2)

Where

(C-3)

The differentials are obtained by use of the cubic spline.

The LHS of Eq. C-2 can be expressed as
(C-4)
SPE-189092-MS 15

Appendix D
Stopping Criteria for Both Levels
When there is a minor change in the objective function

(D-1)

When there is a minor change in the parameters

(D-2)
Where γ1 = 0.01; γ2 = 0.0000001

View publication stats