An intelligent formation pressure prediction method integrating physical and engineering priors

Formation pressure prediction is a critical task in drilling engineering, as its accuracy directly affects well control safety, drilling risk management, and the determination of appropriate drilling fluid density windows. Conventional formation pressure prediction methods mainly rely on a single data source, such as well logs or seismic data. In complex structural settings or abnormal overpressure zones, they often fail to accurately characterize the location and magnitude of pressure variations, leading to considerable prediction uncertainty. For example, in the Huangliu Formation of the Ledong area in the South China Sea, abnormal overpressure is developed, and the associated pressure variations exhibit strong correspondence with seismic impedance and lithological changes. Meanwhile, engineering responses, such as drilling fluid density adjustments, usually indicate abnormal overpressure at shallower depths than the physical variations inferred from seismic or logging data. This phenomenon indicates that engineering information and physical information are complementary in identifying formation pressure anomalies. This study proposes an intelligent formation pressure prediction method that integrates physical and engineering priors. Physical priors derived from seismic inversion and engineering well-control priors are jointly embedded into a deep learning model to achieve accurate pre-drilling pressure prediction for target wells. Application results from field data show that the proposed method improves formation pressure prediction accuracy while effectively capturing pressure variation characteristics. It also provides higher reliability in identifying high-risk intervals and determining pressure safety windows in key formations.