CNNHM

The goal of CNNHM is to provide a package to automatically evaluate the simulation results from the petroleum simulation software instead of manually checking.

The structure of the package

To fluently run the package, we suggest to organize the files in the following way. The frameworks and corresponding versions are also listed below.

├── setup.py
├── CNNHM.py
├── Readme.ipynb
├── training_dataset
│   └── modelA
│   └── modelB
├── validation_dataset
│   ├── GEC1PP99_202_WSWP.png
│   ├── GEC3PP95_202_WSWP.png
│   ├── GEC8PP95_202_WSWP.png
│   └── QAIKPP82_202_WSWP.png
└── saved_model
    ├── modelB_dbscan_100
    ├── modelB_gaussian_100
    └── modelB_kmeans_100

tensorflow                2.0.0
pandas                    1.3.0
numpy                     1.20.3

The python package CNNHM can be reuse in any similar task. You can simply install the package by running the flowing command in your project

python setup.py install

We would like to present the details of the package dependencies here.

import os
from setuptools import setup

setup(
    name = "CNNHM",
    version = "1.0.0",
    author = "Liang Zhang",
    author_email = "liang.zhang@kaust.edu.sa",
    description = ("CNN based evaluator for resevior simulation of history matching"),

    python_requires='>=3',

    install_requires=[
        "tensorflow == 2.0.0",
        "pandas == 1.3.0",
        'numpy ==1.20.3',
    ],
)

Import the package and setup a agent to do the history matching

• dataset_path → which dataset you would like to choose (modelA or modelB)
• label_file → which labelling method you would like to use (kmeans, gaussian, or dbscan)
• geology_property → what geology property you want to evaluate (WSWP, WWCT, WOPR, or WGPR)

from CNNHM import CNN_HM
agent = CNN_HM(dataset_path="/ibex/scratch/zhanl0g/projects/AraIn/PetroleumHM/dataset/modelB",label_file="san_clusterred_modelB_kmeans_WSWP.csv", geology_property="WSWP")

Train

• epochs→ the number of training epochs.
• save_location→ where to save the model (optional).

agent.train(epochs=100,save_location="/ibex/scratch/zhanl0g/projects/AraIn/PetroleumHM/saved_model/modelB_kmeans_100") # if you want to save the model, input the location for saving models

Predict

• prediction_samples → where you put the images you want to know the category they belong to (eg. type1 means overfitting, type 2 means well fitting, type 3 means under fitting, etc.).
• load_model→ if you already train and save the model, can directly load the model from this address, no need to train from the beginning (optional).

agent.predict(**prediction_samples**="/ibex/scratch/zhanl0g/projects/AraIn/PetroleumHM/prediction_samples", load_model="/ibex/scratch/zhanl0g/projects/AraIn/PetroleumHM/saved_model/modelB_kmeans_100")

Output:

Sample 1 QAIKPP82_202_WSWP.png
The prediction result: type 2
The result suppose to be: type 2

Sample 2 GEC3PP95_202_WSWP.png
The prediction result: type 3
The result suppose to be: type 3

Sample 3 GEC8PP95_202_WSWP.png
The prediction result: type 1
The result suppose to be: type 1

Sample 4 GEC1PP99_202_WSWP.png
The prediction result: type 1
The result suppose to be: type 1

Sample 1 QAIKPP82_202_WSWP.png The prediction result: type 2 The result suppose to be: type 2

Sample 2 GEC3PP95_202_WSWP.png The prediction result: type 3 The result suppose to be: type 3

Sample 3 GEC8PP95_202_WSWP.png The prediction result: type 1 The result suppose to be: type 1

Sample 4 GEC1PP99_202_WSWP.png The prediction result: type 1 The result suppose to be: type 1

demo

from CNNHM import CNN_HM
agent = CNN_HM(dataset_path="/Users/liangzhang/Desktop/liang/ibex/projects_ibex/AraIn/PetroleumHM/dataset/modelB",label_file="san_clusterred_modelB_kmeans_WSWP.csv", geology_property="WSWP")
agent.train(epochs=100,save_location="//Users/liangzhang/Desktop/liang/ibex/projects_ibex/AraIn/PetroleumHM/saved_model/modelB_kmeans_100") # if you want to save the model, input the location for saving models
agent.predict(prediction_samples="/Users/liangzhang/Desktop/liang/ibex/projects_ibex/AraIn/PetroleumHM/dataset/modelBprediction_samples", load_model="/Users/liangzhang/Desktop/liang/ibex/projects_ibex/AraIn/PetroleumHM/dataset/modelB/saved_model/modelB_kmeans_100")