Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
Scribd Logo
Upload

Welcome to Scribd!

  • 153 views

    The MPF Theory

    Uploaded by

    krackku k
    1. MPS models facies distributions by considering multiple data points simultaneously rather than just pairwise relationships, allowing it to reproduce complex geological patterns from a training image. 2. The training image defines the conceptual geological model as a facies map that captures spatial relationships between facies. 3. During MPS, a search template is moved over the training image to identify and classify patterns of facies values occurring around unsampled locations. These pattern probabilities guide the simulation. 4. The MPS algorithm works by sequentially visiting all empty nodes and assigning

    Copyright:

    © All Rights Reserved
    Download as PPT, PDF, TXT or read online from Scribd

    The MPF Theory

    Uploaded by

    krackku k
    153 views11 pages
    1. MPS models facies distributions by considering multiple data points simultaneously rather than just pairwise relationships, allowing it to reproduce complex geological patterns from a training image. 2. The training image defines the conceptual geological model as a facies map that captures spatial relationships between facies. 3. During MPS, a search template is moved over the training image to identify and classify patterns of facies values occurring around unsampled locations. These pattern probabilities guide the simulation. 4. The MPS algorithm works by sequentially visiting all empty nodes and assigning

    Original Description:

    Theory of MPF

    Copyright

    © © All Rights Reserved

    Available Formats

    PPT, PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    1. MPS models facies distributions by considering multiple data points simultaneously rather than just pairwise relationships, allowing it to reproduce complex geological patterns from a training image. 2. The training image defines the conceptual geological model as a facies map that captures spatial relationships between facies. 3. During MPS, a search template is moved over the training image to identify and classify patterns of facies values occurring around unsampled locations. These pattern probabilities guide the simulation. 4. The MPS algorithm works by sequentially visiting all empty nodes and assigning

    Copyright:

    © All Rights Reserved
    Download as PPT, PDF, TXT or read online from Scribd
    Download as ppt, pdf, or txt
    153 views11 pages

    The MPF Theory

    Uploaded by

    krackku k
    1. MPS models facies distributions by considering multiple data points simultaneously rather than just pairwise relationships, allowing it to reproduce complex geological patterns from a training image. 2. The training image defines the conceptual geological model as a facies map that captures spatial relationships between facies. 3. During MPS, a search template is moved over the training image to identify and classify patterns of facies values occurring around unsampled locations. These pattern probabilities guide the simulation. 4. The MPS algorithm works by sequentially visiting all empty nodes and assigning

    Copyright:

    © All Rights Reserved
    Download as PPT, PDF, TXT or read online from Scribd
    Download as ppt, pdf, or txt
    You are on page 1of 11

    Multipoint Facies Modeling

    Petrel 2009.1.1

    Why multi point statistics?


    Classical geostatistical methods explain
    only the two-point relationship (variogram)
    In case of only a few hard data (wells)
    connectivity is difficult to quantify
    No direct implementation of geological
    concepts (only via probability properties in
    1D/2D/3D)
    Some facies associations are difficult to
    model using object models
    Multipoint Facies Modeling
    Existing modeling methods

    Two different images in classical geostatistics can be


    based on the same variogram model. The volume might
    be similar but the dynamic model not.

    Object model
    Pixel model

    Variogram Variogram ’replica’


    Multipoint Facies Modeling
    Introduction

    Multi Point Facies Modeling is a stochastic algorithm to


    “convert” a training image (geological conceptual model)
    into a reservoir model.
    Multipoint means to explore the relationship
    between one-to-many points at the same time

    In Multi Point Facies Modeling the variogram will


    be replaced by the Training Image and the MP
    Facies Pattern.
    Multipoint Facies Modeling
    Process
    Sequential multi-point simulation – The local conditional distribution is
    built by directly scanning the training image using a Search mask

    Simulation Grid Search mask Currently simulated


    - Containing 4 data values pixel at location (u)
    - All 4 data + geometry = Pattern
    Training Image
    Channel
    Sand
    Shale

    Pattern
    - Scans the Training image
    --> Infer probability of sand at loc. (u)
    Multipoint Facies Modeling
    Steps

    Multi point facies modeling - is a 3 step procedure


    Training image generation
    Create a simple Geological conceptual model as
    a Petrel 3D property
    Pattern creation
    Applies Search Mask and Multigrid concept
    Uses Neighbourhood Probabilities : :: :: :: :: : : :
    : :: :: :: :: : : :
    Facies modeling
    Use Pattern as main input, condition to
    Hard and soft data, rotation and scaling
    Local update and Well conditioning
    Multipoint Facies Modeling
    Training Image & Pattern

    A Training image is an idealized and simplified representation of the


    geology. It shows the relationship between the different facies. In
    Petrel the training image is a property; derived from:
    Hand drawn images (interactive drawing)
    Existing objects
    Resampled seismic attributes
    Aerial images
    Outcrop analogs

    A Pattern means investigation (scanning) of the neighborhood


    relationship between the facies and then write it into a tree
    Multipoint Facies Modeling
    Training Image & Pattern
    What is the probability of finding the above neighbourhood relationship in a
    model: a) for channel b) for the background

    When counting all possible events where a cell (green) has channel above and
    to the right and background to the left and below there are 7 out of 16
    events showing channel  probability of 0.44 or 44%.
    Multipoint Facies Modeling
    Search mask & Search tree
    Pattern/search tree describes a probability for every possible event. Size depends
    on training image size and search mask. It uses a Multigrid concept for short,
    middle and long distance search

    : : : : : : : : : : : :
    : : : : : : : : : : : :
    Elliptical search mask Rectangular search mask
    2D 3D
    2D 3D

    Petrel stores the pattern in the Input pane. It is not visible to the user.
    Multipoint Facies Modeling
    Training Image & Pattern
    The Multipoint facies simulation uses the pattern as the only mandatory input.
    Additionally one can use:

    Regions (explaining different


    geological situations in different
    areas)
    Soft data (mostly derived from
    seismic; e.g. via Neural Network)
    Hard data (Upscaled cells and
    additional user defined hard data)
    Scaling /Rotation (squeeze,
    stretch and change orientation)
    Multipoint Facies Modeling
    Multi-point facies simulation
    The Multi-point facies simulation setup dialog:

    Regions
    Patterns

    Scaling

    Rotation
    Exercise
    Module 5
    - MPS Theory (2D)

    p. 159 to 182

    You might also like