A hybrid model of formation pore pressure prediction based on geological sequence matching

Formation pore pressure (Pp) is vital to every stage of petroleum exploration and development. However, current prediction methods often overlook geological sequence variations and rely heavily on direct depth alignment from offset wells, resulting in substantial discrepancies between predicted and actual Pp when measured data are scarce. To address these limitations, a novel and interpretable pre-drilling Pp prediction strategy was developed. First, Geological Sequence Matching (GSM) was introduced to align historical well-logging data with the pre-drilling well's stratification, thereby compensating for stratigraphic depth-thickness discrepancies induced by geological evolution—an essential factor systematically neglected in existing approaches. Second, a primary wave velocity (Vp) Error Compensation Hybrid (VECH) model was proposed, which uniquely combines a Vp-based physical model as the primary framework while employing machine learning specifically to correct systematic errors. Unlike purely machine-learning-based or traditional physical methods, VECH maintains robust physical interpretability while effectively incorporating real-world data corrections. By leveraging Vp to calculate effective stress, this approach eliminates the need for post-drilling corrected Pp in model training, overcoming a critical drawback of conventional workflows. Examples from the Bohai Oilfield show that, compared to traditional methods, the proposed hybrid model reduces the mean absolute error in Pp prediction by two-thirds. Furthermore, VECH model interpretation using decision-tree visualization and sensitivity analysis is performed to illustrate the model operation process and the influence of various features on the prediction outcomes. These findings demonstrate the effectiveness of the hybrid model in predicting Pp suggests potential applications in forecasting other geophysical parameters such as density, porosity, and permeability.