A Novel Approach for Evaluating Dominant Factors of Casing Deformation in Deep Shale Gas Wells Based on SHAP Values

As an unconventional energy resource, shale gas plays a critical role in the energy structure. In China, the deep shale gas formations are characterized by high stress and high complexity, which lead to extremely severe casing deformation issues, causing many wells to be abandoned. This not only hinders the growth of Estimated Ultimate Recovery (EUR), but also increases the operational cost and difficulty, shortens the wellbore life cycle, and reduces the overall economic benefits of shale gas development. Therefore, this study presents an interpretable machine learning (ML) framework combining Random Forest (RF), XGBoost, and LightGBM models with SHapley Additive exPlanations (SHAP) to evaluate key geological factors influencing casing deformation. Preprocessed geological data (e.g., porosity, stress difference, fracturing time) was used to train and validate the ML models. The RF model achieved the highest performance (accuracy: 92.76%, AUC: 0.954). SHAP analysis identified porosity, minimum horizontal stress, and Young’s modulus as the dominant casing deformation factors. Critical thresholds were established: porosity >4.5% and Young’s modulus 40,000–42,000 MPa substantially increase casing deformation risk, while a stress difference of 11–13 MPa and maximum horizontal stress <110 MPa reduce it. These findings, consistent with field observations and geomechanical principles, enhance casing deformation prediction accuracy and transparency. The proposed method aids in mitigating casing deformation risks, improving EUR, and promoting sustainable shale gas development.