Method for automatically extracting image features of electrical imaging well logging, computer equipment and non-transitory computer readable medium

Patent metadata

Field	Value
Publication number	US-11010629-B2
Application number	US-201916507270-A
Country	US
Kind code	B2
Filing date	Jul 10, 2019
Priority date	Aug 24, 2018
Publication date	May 18, 2021
Grant date	May 18, 2021

Abstract

A method and an apparatus for automatically extracting image features of electrical imaging well logging, wherein the method comprises the steps of: acquiring historical data of electrical imaging well logging; pre-processing the historical data of the electrical imaging well logging to generate an electrical imaging well logging image covering a full hole; recognizing and marking a typical geological feature in the electrical imaging well logging image covering the full hole, obtaining a processed image, and determining the processed image as a training sample according to types of the geological features; constructing a deep learning model including an input layer, a plurality of hidden layers, and an output layer; training the deep learning model using the training sample; using the trained deep learning model, recognizing type of a geological feature of an electrical imaging well logging image of a well section to be recognized, and performing morphological optimization processing on the recognition result to obtain a feature optimization recognition result. The solution can automatically, quickly and accurately recognize the typical geological features in the electrical imaging well logging image.

First claim

The invention claimed is:

1. A method for automatically extracting image features of electrical imaging well logging, wherein the method comprises:
acquiring historical data of electrical imaging well logging;
pre-processing the historical data of the electrical imaging well logging to generate an electrical imaging well logging image covering a full hole;
recognizing and marking a typical geological feature in the electrical imaging well logging image covering the full hole, obtaining a processed image, and determining the processed image as a training sample according to types of the geological features;
constructing a deep learning model including an input layer, a plurality of hidden layers, and an output layer;
training the deep learning model using the training sample to obtain a trained deep learning model;
using the trained deep learning model, recognizing type of a geological feature of an electrical imaging well logging image of a well section to be recognized to obtain a recognition result; and
performing morphological optimization processing on the recognition result to obtain a feature optimization recognition result;
wherein the deep learning model is constructed as follows:
the deep learning model includes a twelve-layer structure comprising: a first layer, a second layer, a third layer, a fourth layer, a fifth layer, a sixth layer, a seventh layer, an eighth layer, a ninth layer, a tenth layer, an eleventh layer, and a twelfth layer;
a structure of each layer of the twelve-layer structure is as follows:
the first layer is an input layer;
the second layer is a hidden layer, including a convolution layer, an activation function layer, a convolution layer, an activation layer, and a pooling layer;
the third layer is a hidden layer, including a convolution layer, an activation function layer, a convolution layer, an activation layer, and a pooling layer;
the fourth layer is a hidden layer, including a convolution layer, an activation function layer, a convolution layer, an activation layer, a convolution layer, an activation layer, and a pooling layer;
the fifth layer is a hidden layer, including a convolution layer, an activation function layer, a convolution layer, an activation layer, a convolution layer, an activation layer, and a pooling layer;
the sixth layer is a hidden layer, including a convolution layer, an activation function layer, a convolution layer, an activation layer, a convolution layer, an activation layer, and a pooling layer;
the seventh layer is a hidden layer, including a full convolution layer, an activation function layer, and a Dropoout layer;
the eighth layer is a hidden layer, including a full convolution layer, an activation function layer, and a Dropoout layer;
the ninth layer is a hidden layer, including a deconvolution layer, a Crop layer, and an Eltwise layer;
the tenth layer is a hidden layer, including a deconvolution layer, a Crop layer, and an Eltwise layer;
the eleventh layer is a hidden layer, including a deconvolution layer and a Crop layer; and
the twelfth layer is an output layer, including a softmax layer.

2. The method for automatically extracting image features of electrical imaging well logging according to claim 1 , wherein pre-processing the historical data of the electrical imaging well logging to generate an electrical imaging well logging image covering a full hole includes:
performing acceleration correction processing and equalization processing on the historical data of the electrical imaging well logging to obtain an original electrical imaging well logging image;
performing resistivity scale processing on the original electrical imaging well logging image to obtain a scale image reflecting resistivity of rocks of borehole wall formation; and
performing full hole image generation processing on the scale image reflecting resistivity of rocks of borehole wall formation, to generate an electrical imaging well logging image covering the full hole.

3. The method for automatically extracting image features of electrical imaging well logging according to claim 2 , wherein recognizing the well logging facies of the electrical imaging well logging image of the well section to be recognized using the trained deep learning model includes:
performing acceleration correction processing and equalization processing, resistivity scale processing, and a full hole image generation process on the electrical imaging well logging image of the well section to be recognized, to generate an electrical imaging well logging image covering the full hole that is to be recognized;
performing depth-by-depth window processing on the electrical imaging well logging image covering the full hole that is to be recognized, to obtain a plurality of depth window images; and
inputting the plurality of depth window images into the trained deep learning model to obtain a recognition result.

4. The method for automatically extracting image features of electrical imaging well logging according to claim 1 , wherein the method includes the step of:
performing statistics on geological feature development parameters of the recognition result according to the geological feature type; and
evaluating a development degree of a reservoir according to the statistical geological feature development parameters.

5. The method for automatically extracting image features of electrical imaging well logging according to claim 4 , wherein the geological feature development parameters include one or more of a surface porosities of development of different geological features, grain diameters of different geological features and number of development of different geological features.

6. A computer device comprising:
a memory,
a processor, and
a computer program stored on the memory and executable on the processor, wherein when executing the computer program, the processor implementing the following:
acquiring historical data of electrical imaging well logging;
pre-processing the historical data of the electrical imaging well logging to generate an electrical imaging well logging image covering a full hole;
recognizing and marking a typical geological feature in the electrical imaging well logging image covering the full hole, obtaining a processed image, and determining the processed image as a training sample according to types of the geological features;
constructing a deep learning model including an input layer, a plurality of hidden layers, and an output layer;
training the deep learning model using the training sample to obtain a trained deep learning model;
using the trained deep learning model, recognizing type of a geological feature of an electrical imaging well logging image of a well section to be recognized to obtain a recognition result; and
performing morphological optimization processing on the recognition result to obtain a feature optimization recognition result;

wherein the deep learning model is constructed as follows:
the deep learning model includes a twelve-layer structure comprising: a first layer, a second layer, a third layer, a fourth layer, a fifth layer, a sixth layer, a seventh layer, an eighth layer, a ninth layer, a tenth layer, an eleventh layer, and a twelfth layer;
a structure of each layer of the twelve-layer structure is as follows:
the first layer is an input layer;
the second layer is a hidden layer, including a convolution layer, an activation function layer, a convolution layer, an activation layer, and a pooling layer;
the third layer is a hidden layer, including a convolution layer, an activation function layer, a convolution layer, an activation layer, and a pooling layer;
the fourth layer is a hidden layer, including a convolution layer, an activation function layer, a convolution layer,

Assignees

• Petrochina Co Ltd

Inventors

• Feng Zhou
• Li Ning
• Wu Hongliang
• Wang Kewen
• Liu Peng
• Li Yusheng
• Wang Huafeng
• Wang Chen

Classifications

• G06V10/454 Primary

  Integrating the filters into a hierarchical structure, e.g. convolutional neural networks [CNN] · CPC title

• G06V10/82

  using neural networks · CPC title

• G06V10/764

  using classification, e.g. of video objects · CPC title

• G06N3/045

  Combinations of networks · CPC title

• G06N3/048

  Activation functions · CPC title