CN-121998134-A - Face rate prediction method, device, equipment and storage medium based on random forest method

CN121998134ACN 121998134 ACN121998134 ACN 121998134ACN-121998134-A

Abstract

The invention discloses a face rate prediction method, device and equipment of a random forest method and a storage medium, and relates to the technical field of unconventional oil and gas exploration and development. The invention relates to a machine learning method based on random forests, which utilizes logging data and core data to fit face rate parameters (including organic face rate, inorganic face rate and crack face rate) of shale, trains a face rate RF model, predicts the face rate data of other wells in a research block through the trained face rate RF model, provides a convenient means for identifying desserts of the shale, reduces the acquisition cost of the face rate parameters, and simplifies the acquisition process of the face rate parameters.

Inventors

LI YIZHEN
CHEN XUE
Hao Yuexiang
JIANG YUMENG
He Huaiyin
YANG YADONG
JIANG WEI
ZHOU XIN
QIAN CHAO
LIU YI

Assignees

中国石油天然气集团有限公司
中国石油集团川庆钻探工程有限公司

Dates

Publication Date: 20260508
Application Date: 20241104

Claims (15)

1. A face rate prediction method based on a random forest method is characterized by comprising the following steps: s1, collecting logging curves and core electron microscope scanning data of all wells in a research block; S2, respectively aligning the core electron microscope data of each well with the depth of the corresponding logging data, wherein one part of the aligned data is used as a training sample set of the training surface porosity RF model, and the other part is used as a verification sample set of the trained surface porosity RF model; s3, carrying out correlation analysis on the data in the training sample set in the step S2, and respectively screening out curves with higher correlation corresponding to the face rate data; S4, dividing the screened curve into a training set and a testing set of the training surface porosity RF model; S5, training the surface porosity RF model by using a training set, testing the trained surface porosity RF model by using a testing set, and checking the reliability of the surface porosity RF model, if the deviation between the predicted result and the actual result is large, readjusting parameters until the trained surface porosity RF model passes the reliability check; s6, performing model verification on the face rate RF model obtained through training in the step S5 by using the verification sample set in the step S2, if the verification is passed, the face rate RF model can be applied to the research block, otherwise, the step S3-S5 is performed again; s7, predicting the surface area rate data of other wells of the research area by using the surface area rate RF model verified in the step S6.
2. The method for predicting the surface porosity based on the random forest method according to claim 1, wherein the surface porosity comprises an organic surface porosity, an inorganic surface porosity and a crack surface porosity, the organic surface porosity RF model, the inorganic surface porosity RF model and the crack surface porosity RF model are obtained through training and testing of an RF model and are applied to other wells of a research block to respectively predict the organic surface porosity, the inorganic surface porosity and the crack surface porosity of the other wells of the research block.
3. The method for predicting the face rate based on the random forest method according to claim 1 or 2, wherein in the step S1, the logging data comprises a logging curve of a target layer and mineral model data obtained by calculation.
4. The method for predicting the face rate based on the random forest method of claim 1 or 2, wherein in the step S2, the depth of the core electron microscope data is used as a standard, and the logging data and the core electron microscope data are aligned.
5. The method of predicting face rate based on random forest method of claim 4, wherein in step S2, the data obtained by deep alignment of the core electron microscope data and the logging data is exported to form a file.
6. The face rate prediction method based on the random forest method according to claim 1 or 2, wherein in the step S4, the dividing ratio of the training set to the testing set is 0.6-0.9:0.4-0.1, and the sum of the training set ratio and the testing set ratio is not more than 1.
7. The method for predicting the face rate based on the random forest method of claim 6, wherein in the step S4, the dividing ratio of the training set to the testing set is 0.7:0.3.
8. The method for predicting the face rate based on the random forest method according to claim 1 or 2, wherein in the step S5, the face rate RF model is trained by a method of adjusting parameters, and the adjusted parameters comprise the number of feature trees, the number of random seeds and the number of feature leaves.
9. The method of predicting face rate based on random forest method of claim 8, wherein in step S5, the reliability of the RF model of face rate is checked, and the parameters are readjusted if the predicted result deviates greatly from the actual result.
10. The method for predicting the face rate based on the random forest method according to claim 1 or 2, wherein in the step S6, before the data in the verification sample set is subjected to model verification on the face rate RF model, the curve number and the curve sequence of the data input into the face rate RF model in the verification sample set are corresponding to the training set when the model is built.
11. The method for predicting the surface porosity based on the random forest method of claim 10, wherein in the step S6, the surface porosity predicted by using logging data is compared with the surface porosity obtained by core electron microscope scanning data in the process of model verification, if the difference is not large, the model is accurate and can be applied to the research block, otherwise, if the difference is large, the model is not accurate to build, and the surface porosity RF model needs to be trained again according to the steps S3-S5.
12. The method for predicting the surface area rate based on the random forest method according to claim 1 or 2, wherein in the step S7, the number of curves and the order of the curves of the logging data of other wells input when the surface area rate data of other wells of the research area are predicted by using the verified surface area rate RF model in the step S6 correspond to the training set when the model is built.
13. A face rate prediction device based on a random forest method is characterized in that the prediction device comprises, The model data collection module is used for collecting logging curves and core electron microscope scanning data of all wells in the research block and respectively aligning the core electron microscope data of all wells with the corresponding logging data depth; The model data preprocessing module is used for carrying out correlation analysis on the data after depth alignment and respectively screening out curves with higher correlation corresponding to the face rate data; The model building module is used for training the surface porosity RF model according to the curve with higher correlation corresponding to the surface porosity data, carrying out reliability check and accuracy verification on the trained surface porosity RF model, and retraining the surface porosity RF model by adjusting parameters if the reliability check is not passed; and the data prediction module is used for substituting the surface porosity RF model trained by the model building module into logging data of other wells in the research block to predict the surface porosity of the other wells.
14. A computer device comprising a processor, an input device, an output device and a memory, the processor, the input device, the output device and the memory being interconnected, wherein the memory is adapted to store a computer program comprising program instructions, the processor being configured to invoke the program instructions to perform part or all of the steps of a random forest method based face rate prediction method as claimed in any of claims 1-12.
15. A computer-readable storage medium, characterized in that the computer-readable storage medium stores a computer program for electronic data exchange, wherein the computer program causes a computer to execute part or all of the steps of a face rate prediction method based on a random forest method according to any one of claims 1 to 12.

Description

Face rate prediction method, device, equipment and storage medium based on random forest method Technical Field The invention relates to the technical field of unconventional oil and gas exploration and development, in particular to a face rate prediction method, device and equipment based on a random forest method and a storage medium. Background The dessert evaluation is an important content in the conventional oil and gas exploration and development at present, and has important significance on the development of the benefits of the unconventional oil and gas scale. The concept meaning of dessert is continuously expanded, and the selection and value standard of evaluation parameters are more diversified and regional. According to the dessert classification criteria of a certain region, it is required to classify the dessert by using three face porosity parameters of inorganic face porosity, organic face porosity and crack face porosity as shown in the following table 1 dessert classification criteria. Table 1 shows the dessert partitioning criteria Three surface area ratio parameters in the dessert dividing standard are generally obtained through a digital rock core method, and a series of steps such as taking out a rock core and constructing the digital rock core are needed, so that the cost is high and the steps are complicated. Disclosure of Invention In order to overcome the defects and shortcomings in the prior art, the invention provides a face rate prediction method, a device, equipment and a storage medium based on a random forest method. The invention relates to a machine learning method based on random forests, which utilizes logging data and core data to fit face rate parameters (including organic face rate, inorganic face rate and crack face rate) of shale, provides a convenient means for identifying desserts of the shale, reduces the acquisition cost of the face rate parameters and simplifies the acquisition process of the face rate parameters. In order to solve the problems in the prior art, the invention is realized by the following technical scheme. The first aspect of the invention provides a face rate prediction method based on a random forest method, which comprises the following steps: s1, collecting logging curves and core electron microscope scanning data of all wells in a research block; S2, respectively aligning the core electron microscope data of each well with the depth of the corresponding logging data, wherein one part of the aligned data is used as a training sample set of the training surface porosity RF model, and the other part is used as a verification sample set of the trained surface porosity RF model; s3, carrying out correlation analysis on the data in the training sample set in the step S2, and respectively screening out curves with higher correlation corresponding to the face rate data; S4, dividing the screened curve into a training set and a testing set of the training surface porosity RF model; S5, training the surface porosity RF model by using a training set, testing the trained surface porosity RF model by using a testing set, and checking the reliability of the surface porosity RF model, if the deviation between the predicted result and the actual result is large, readjusting parameters until the trained surface porosity RF model passes the reliability check; s6, performing model verification on the face rate RF model obtained through training in the step S5 by using the verification sample set in the step S2, if the verification is passed, the face rate RF model can be applied to the research block, otherwise, the step S3-S5 is performed again; s7, predicting the surface area rate data of other wells of the research area by using the surface area rate RF model verified in the step S6. Further preferably, the surface porosities comprise organic surface porosities, inorganic surface porosities and crack surface porosities, and the organic surface porosities, the inorganic surface porosities and the crack surface porosities are obtained through the training and the testing of the RF model, and are applied to other wells of the research block to respectively predict the organic surface porosities, the inorganic surface porosities and the crack surface porosities of other wells of the research block. Further preferably, in step S1, the logging data includes a logging curve of the target layer and calculated mineral model data. Further preferably, in step S2, the logging data and the core electron microscope data are aligned with each other by using the depth of the core electron microscope data as a standard. Still more preferably, in step S2, the data obtained by depth alignment of the core electron microscope data and the logging data is exported to form a file. Further preferably, in the step S4, the dividing ratio of the training set to the testing set is 0.6-0.9:0.4-0.1, and the sum of the training set ratio and the testing set ratio is not more than 1. Sti