Research On Coalbed Methane Production Data Mining And Artificial Intelligent Reservoir Dynamic Analysis

In recent years,data mining and artificial intelligence technology have developed rapidly,and the application in the research in the coalbed methane(CBM)industry has been gradually popularized.However,the application process is generally faced with the following problems: the existing algorithms are often directly transplanted into CBM research from specific fields,so the application effect is poor.It is difficult to explain the mechanism by directly applying machine learning algorithms and abandoning the previous mechanism research results.The application of machine learning is too limited to gas production prediction models and lacks the application research of artificial intelligence.Therefore,it is very important to study how to organically integrate advanced data science algorithms and artificial intelligence technology into CBM theoretical research,increase mechanism understanding,broaden application scope and realize real intelligence.Starting from CBM related theories such as gas reservoir engineering,gas lifting engineering,and seepage mechanics,this paper explores the integration method of data mining,artificial intelligence technology,and CBM theory.The main research contents and conclusions are as follows:(1)Aiming at the lack of targeted data preprocessing technology in the CBM industry,the outlier identification and denoising method suitable for CBM production data set are studied.By analyzing the generation mechanism of abnormal value data in CBM production data,the abnormal value is classified.The automatic identification technology of abnormal value of CBM production data is formed combined with the division of drainage and production stage and the adaptive method judgment threshold.By establishing the interpolation models of the production system and pressure system,combined with the matrix decomposition method,the noise removal method of CBM production data is formed.Through the establishment of outlier seed detection method,the actual production data are tested to verify the feasibility of the technology.(2)Aiming at the limitations of the current research direction of data mining in CBM industry and the difficulty to explain problems in the mechanism research,the application mode of data mining in the field of CBM is explored.Firstly,aiming at the problem that the current data mining is mostly limited to productivity prediction,the application mode of data mining in CBM production data association mining is explored.Taking the relationship between the source of reservoir damage and the connotation of drainage management and control as the research object,combined with the theory of coalbed water lock damage,stress sensitivity,and CBM damage caused by pulverized coal subsidence,the source of main reservoir damage is inversed by numerical simulation method.The correlation between the main damage of CBM reservoir and coal reservoir physical property and drainage and production system is excavated by using characteristic correlation algorithm.Secondly,aiming at the problem that the data mining algorithm abandons the traditional research results of CBM mechanism and the model is difficult to explain,the combination of theoretical model and machine learning model is explored.Taking the wellbore pressure model as the research object,referring to the modeling idea of deep learning algorithm and random gradient descent optimization algorithm,the gas reservoir engineering topology network modeling method is proposed,and the wellbore pressure system topology network model is established to improve the calculation accuracy of bottom hole flow pressure and enhance the interpretability of the network model.(3)Aiming at the problem of how to integrate artificial intelligence into gas reservoir engineering research,the artificial intelligence reservoir dynamic analysis method is studied.Firstly,a CBM numerical simulator using GPU is built.Based on the deep reinforcement learning algorithm,the artificial intelligence automatic history matching technology is achieved.The single parameter dynamic analysis method is realized by building the reservoir physical property inversion agent.Then,by constructing the characterization model of fracturing fracture effectiveness,combined with artificial intelligence automatic history matching technology,the artificial intelligence reservoir dynamic analysis technology of fracturing effect correction and fracturing effectiveness dynamic tracking is realized.(4)Aiming at the quantitative optimization of CBM drainage and production control,the intelligent drainage and production analysis system of CBM is studied.Combined with the research results of(1)(2)(3),through the division of drainage and production stages,flow pressure control pattern recognition,inter well interference prediction,reinforcement learning algorithm,the drainage and production control decision-making agent is established,and an integrated CBM intelligent drainage and production analysis system is formed finally.The functions of systematic analysis of the causes of inefficient wells and quantitative optimization of drainage and production management and control are realized,which provides technical support for reasonably improving the productivity of CBM wells.