C Z Zhang, Xuehai Fu, Zhanxin Xue, Chaochao Duan, Kexin Li, Xiaofeng Hao, Huimin Jia, Guangjie Sang, Zeqing Lei, Shijie Yu, Baoxin Zhang
Geological engineering differences generally lead to strong heterogeneity in coalbed methane (CBM) production. More than 800 CBM wells are in production in the MB area in the southern Qinshui Basin, North China, and differences in gas production between wells reach 10 times. However, mechanisms of productivity variations remain unclear. Based on drilling, logging, and production data, this study employed machine learning methods to identify key factors controlling gas production. The results showed that the peak gas production of vertical wells with No. 3 and No. 15 coal co-production was generally below 5000 m3/d, and that of most horizontal wells exceeded 10,000 m3/d, especially higher in wells for No. 15 coal. High-production wells were concentrated in the south-eastern regions, predominantly accompanied by high gas content and a critical desorption pressure ratio. Based on the LightGBM+SHAP model analysis, three distinct production control models were established: the critical desorption pressure ratio-dominated and fracturing scale synergy model (horizontal wells for No. 3 coal), the permeability-dominated with reservoir pressure and critical desorption pressure ratio support model (horizontal wells for No. 15 coal), and the average critical desorption pressure ratio-dominated and total sand injection volume-supported model (vertical wells with No. 3 and No. 15 coal co-production). These findings provide an approach for identifying key parameters affecting gas production, and offer insights for formulating efficient CBM development strategies.