水力压裂深层储层破裂压力预测可解释人工智能模型OA

To address the limited interpretability of existing models for predicting fracture pressure during hydraulic fracturing in deep reservoirs,this paper develops a multi-granularity interpretable framework integrating Shapley additive explanations(SHAP)and Kolmogorov-Amold network(KAN)based on field engineering data.The proposed framework achieves high prediction accuracy with an average coefficient of determination of 0.958 5.Interpretability analysis reveals that at the macro level,torque-related features dominate the prediction process.At the micro level,engineering parameters such as weight on bit and torque control fracture pressure by affecting rock tensile strength and in-situ stress levels,with stress level identified as the dominant controlling factor.These findings are consistent with existing theoretical and experimental studies.The macro-level feature attribution from SHAP and the micro-level mechanism interpretation from KAN mutually validate each other,demonstrating the synergistic advantages of the proposed multi-granularity interpretable framework.