Prediction Of Coalbed Gas Content Based On Well Logging Data And CNN-GRU Model

Gas content is one of the key parameters for coal reservoir evaluation,and it is an important basis for coalbed methane resource exploration and development or production.Accurate evaluation of gas content is important for the identification and efficient development of favorable development blocks for coalbed methane resources.With the development and application of physical logging technology,using well logging data to predict the gas content of coal seams has been proven to be an effective and feasible solution.Based on the logging data,the deep learning method is applied to excavate the nonlinear implicit relationship between the logging data and the gas content,so as to realize the prediction of the gas content,which has theoretical significance and practical value for the development of coalbed methane.Based on CNN-GRU deep learning model,we take Fanzhuang-Zhengzhuang logging data and gas-bearing coring experimental data in Qinshui Basin as samples to predict the coal seam gas content On the basis of analysis of response characteristics and sensitivity of coal seam gas content logging combined with clustering algorithm.The main work and the understanding obtained are as follows:To address the problems of poor quality of logging data and distorted or missing logging curves,the BGRU model is used to make full use of the sequence information in the logging data based on the correlation between the target logging curves,and the impact of the upper and lower reservoirs on the missing points is considered comprehensively to achieve the task of logging curve reconstruction,and the prediction accuracy can reach 85%.Feature analysis,extraction and clustering of log data.The response characteristics of the logging attributes to the gas content of the coal seam were analyzed by the rendezvous plot method.In view of the inconsistency between the gas content and the response of logging parameters,K-means clustering analysis of logging curves is used to classify the reservoir categories.Based on the gas content-logging parameter response analysis and clustering results,a CNN-GRU neural network model was constructed to predict the gas content and validate the application.Using the logging data of No.15 coal seam of Taiyuan Group in the study area,we extracted logging parameter features,analyzed the gas content-logging parameter response features,and built a CNN-GRU neural network model,and used the logging data of No.3 coal seam of Shanxi Group for gas content prediction and application validation.Based on the main line of "logging curve reconstruction,gas content logging response feature analysis and logging parameter selection,neural network model training and application validation",a coal seam gas content prediction model based on logging data and deep learning method is constructed.In conclusion,this thesis effectively combines deep learning technology,logging response analysis and geophysical logging data to fully explore the potential relationship between logging data and coal-bed gas content,and provides a methodological reference for coal-bed gas content evaluation and coal-bed methane exploration and development.This Thesis contains 55 pictures,18 tables and 103 references.