Prediction Method of Formation Pore Pressure Driven by the Coupling of Depth Learning and Eaton Method

Accurately predicting formation pore pressure is crucial for ensuring safe drilling construction, but it has remained a longstanding challenge. Traditionally, geophysical logging data is used to evaluate formation pressure, but this method is limited by the subjective and uncertain selection of model parameters. In this paper, the combination of deep learning and the traditional Eaton model is used to establish a more effective and objective method for predicting formation pore pressure by establishing a complex relationship between logged seismic and formation pressure. The paper investigates three prediction methods: the traditional stratified Eaton index method, data-driven formation pore pressure prediction, and LSTM’s Eaton index prediction. The results show that the method established in this paper has an average absolute error percentage of 3.014% for predicting pore pressure in deep complex strata. In comparison, the BP neural network and stratified Eaton index averaging methods have errors of 14.398% and 14.447%, respectively. The joint use of deep learning and the Eaton method not only improves the accuracy of pore pressure prediction in deep complex formations but also incorporates the multi-source data response characteristics of traditional methods. This provides theoretical support for data-driven methods and enhances our understanding of formation pore pressure.