基于SSA-CNN模型的煤储层含气量预测方法研究OA

Accurate prediction of gas content in deep coal reservoirs is of significant engineering value for the efficient development of coalbed methane.However,reliance on a single geophysical parameter often overlooks the control of internal structures and heterogeneity of coal reservoirs on gas content,leading to deviations between predicted results and measured values,which fail to meet the precision requirements for efficient deep coalbed methane extraction.Based on in-depth mining of geophysical information,this study introduces an intrinsic mechanistic parameter of coal reservoirs-ash content-as a supplement to geological characteristics,thereby compensating for the inadequacy of geophysical information and constructing a multimodal feature system for gas content prediction.Taking the No.8 coal reservoir of the Carboniferous Benxi Formation in Block M on the eastern margin of the Ordos Basin as the study object,a convolutional neural network optimized by the Sparrow Search Algorithm(SSA-CNN)was employed to automatically extract spatial features of the data and establish a high-precision gas content prediction model.The results indicate that:1)Through Spearman nonlinear correlation analysis,six factors with significant influence on gas content were selected and determined as inputs for the prediction model;2)The SSA-CNN model achieved a prediction accuracy of R2=0.817 on the test set,with the mean absolute error reduced by 1.067％compared to the traditional CNN model.Practical research and analysis demonstrate that the SSA-CNN model can be effectively applied to high-precision gas content prediction in coal reservoirs and has promotional value for gas content prediction in regions with similar geological backgrounds.