Rock consolidation degree-based, multi-parameter smooth threshold regression method for predicting hydrocarbon-generation overpressure-induced pore pressure

Abnormal pore pressure is a primary cause of drilling-related problems, and accurate pre-drilling prediction of formation pore pressure has become a critical technical component in deepwater oil and gas exploration and development. Prediction accuracy directly affects drilling safety and operational efficiency. Existing methods based on a single parameter (e.g., P-wave velocity) are insufficient to fully characterize the coupled effects among multiple geological parameters, resulting in limited predictive performance. To address this issue, this study proposes a multi-parameter collaborative pore pressure prediction model based on smooth threshold regression. Multiple parameters associated with formation pore pressure are incorporated into the model. First, an optimal feature set is identified using a hybrid feature selection strategy that combines the Lasso algorithm and bidirectional stepwise regression. Subsequently, pore pressure is predicted using the smooth threshold regression model. To remedy the absence of shear-wave data in logging records, a shear-wave prediction method based on an improved consolidation index is introduced. A case study in a hydrocarbon-generating overpressure zone in the Bohai Sea demonstrates that the proposed model achieves a mean absolute error of approximately 1.61 and a root mean square error of approximately 2.02 in the test well, with an average cross-validated coefficient of determination (R2) of 0.9454. The model accurately identifies abnormally overpressured intervals and quantitatively estimates pore pressure, providing reliable technical support for drilling risk mitigation and reservoir evaluation.