Hybrid Model Based on Copula Mutual Information and SSA-BP: Analysis of Key Factors and Prediction of Stable Gas Production

Identifying and predicting the primary factors influencing coalbed methane (CBM) well yield are pivotal for effective fracturing decisions. However, deciphering the nonlinear relationships between CBM fracturing effectiveness and its influencing factors poses significant challenges at the mechanistic level. This paper introduces a novel framework for analyzing key factors and predicting CBM production. Firstly, an enhanced adaptive copula-based feature selection algorithm, which utilizes maximum correlation and minimum redundancy criteria, identifies four key control factors: gas saturation, gas content, pre-liquid volume and sand carrying liquid volume. Then, a hybrid optimization algorithm, merging the salp swarm algorithm and backpropagation neural network, establishes a CBM well production prediction model achieving an 83% coefficient of determination (R²). These results inform the analysis of key control factors in typical block wells, guiding a CO₂ foam fracturing scheme. Following field application, daily gas production escalated from 322 to 950 m³/d, significantly enhancing fracturing effectiveness.