Shale Gas Fracturing Construction Parameter Optimization Based on SSA-XGBoost

Abstract

To address the challenges of poor model performance, low prediction accuracy, and suboptimal application of intelligent optimization algorithms for hydraulic fracturing parameter optimization in shale gas production, this study proposes a novel optimization method based on the SSA-XGBoost algorithm. First, using production data from 120 wells with complete records in the Y block, grey relational analysis (GRA) was employed to reduce dimensionality among 14 factors affecting fracturing performance, resulting in the selection of nine primary controlling factors. Based on the mapping relationship between these controlling factors and the fracturing performance evaluation index (daily gas production), the Sparrow Search Algorithm (SSA) was applied to optimize the hyperparameters of the Extreme Gradient Boosting (XGBoost) model. The resulting SSA-XGBoost-based shale gas production prediction model improved the coefficient of determination (R2) by 5.6% compared to the original model, achieving significantly enhanced prediction accuracy and outperforming existing models in the field. Finally, based on the production predictions of the optimized model, the optimal parameter ranges for hydraulic fracturing were determined, yielding normalized daily gas production 27.50% higher than that of benchmark wells. These findings demonstrate the rationality and accuracy of the proposed optimization approach, providing valuable reference for on-site fracturing parameter design.