Coalbed Methane Content Prediction with a Seismic Multi-attribute Support Vector Regression Model

Abstract

Accurate prediction of coalbed methane (CBM) content plays an essential role in CBM exploration and development. In this study, we selected eight seismic attributes with good responses to the CBM content as the input data. The support vector regression (SVR) model was employed to predict the CBM content and compared with the results of the traditional BP neural network method. The results reveal that the SVR model has higher accuracy compared to the BP neural network model and can better identify areas with high CBM content in the case of small samples. Last, we applied the seismic multi-attribute SVR model to predict the CBM content in an exploration area of the Qinshui Basin in China. The predicted high-gas-bearing areas are consistent with the field data, further verifying the effectiveness and practicality of our method for predicting CBM content.