A Research On Productivity Classification And Intelligent Evaluation Of Coalbed Methane Reservoirs In Shizhuangnan Area,Qinshui Basin

The Qinshui Basin is the first to achieve large-scale development among the nine major coalbed methane basins(groups)in China.As the main production area of the coalbed methane industry in the Qinshui Basin,the overall production of the Shizhuangnan Area is not satisfactory during the past ten years of development,and there are a lagre number of old wells with low production and large production differences.A full understanding of the drainage rules,production conditions and the influencing factors of production capacity in the Shizhuangnan Area is of great significance for finding the remaining sweets,secondary transformation and development,and increasing production.Using the experimental data,drainage data,well logging data and geological data of the Shizhuangnan 3# coalbed,taking the productivity of the production well as the constraint,evaluating of coalbed characteristic parameters and main control factor analysis,and comprehensive consideration of a variety of factors affecting production capacity,the optimized random forest algorithm realizes the productivity classification and intelligent evaluation of Shizhuangnan Area.Based on logging data,the random forest evaluation model,BP neural network evaluation model,and support vector machine model for key characteristic parameters of coalbed,such as industrial composition,coal structure,gas content,critical desorption pressure,Langmuir parameters,and permeability,are established.The multiple regression evaluation model of some parameters has also been realized.It is found that the random forest has the best performance in the evaluation of various parameters.On the basis of the traditional random forest model,it was improved by optimizing the train set,and the optimized model was applied to other production wells and achieved good results.According to the average daily gas production of a single well of CBM wells,the development wells are divided into three types: high-production wells,middle-production wells,and low-production wells.Based on the water production,gas production,key time nodes and bottom hole pressure,the drainage characteristics of the wells of the three productivity levels are analyzed.By summarizing the production characteristics of CBM wells of different productivity levels,the understanding of CBM productivity has been further strengthened.By combining with the optimized characteristic parameters and the classification results of the production wells of the three productivity levels,the main control factors are analyzed from the coal seam,gas content,thickness,permeability,coal structure,industrial composition,adsorption and desorption capacity,roof and floor lithology and buried depth.The optimized random forest algorithm is used to establish a classification method for production wells,which realizes the transformation from relying on experience to data-driven production capacity.The established productivity classification model has a coincidence rate of 97.3% in the study area,which verifies the reliability of the algorithm model and is of great significance for the subsequent evaluation of CBM wells and adjustment of development plans.