Prediction model of coalbed methane content based on well logging parameter optimization

With the development of coalbed methane(CBM) exploration, higher accuracy of CBM content prediction is required. Based on the response characteristics of CBM logging, the correlation between logging parameters and gas content is analyzed, and the optimization strategy of logging parameters by combining MIV technology with LSSVM is proposed. The optimal logging parameters are selected as the input independent variables of network modeling, and the core parameters of LSSVM(Least Squares Support Vector Machine) network are optimized by particle swarm optimization. Finally, a set of MIV-PSO-LSSVM model suitable for CBM content prediction is constructed. The prediction performances of LSSVM, PSO-LSSVM, MIV-LSSVM, MIV-PSO-LSSVM and traditional multiple regression method are compared and analyzed respectively. The results show that the prediction accuracy of LSSVM model optimized by PSO is increased, and the modeling performance of neural network is improved significantly with MIV method to optimize logging parameters. MIV-PSO-LSSVM model could realize high-precision prediction of CBM content, providing new technical support for CBM exploration and reservoir evaluation. And the modeling strategy of this research can be widely used in other ML(machine learning) modeling research fields.