Real-Time Prediction of Formation Pore Pressure Using Hybrid LSTM-BP Model

Abstract

ABSTRACT Accurate and real-time prediction of formation pore pressure is critical to ensure drilling safety. However, the performance of the most traditional formation pore pressure calculation methods are not satisfied in field applications due to their limited accuracy and time-effectiveness. In this paper, a novel artificial intelligence model is proposed to accurately predict formation pore pressure in real-time. First, the dataset for model training is prepared after handling the outliers and missing values of the drilling and logging data, which collected from 12 wells in Tarim Basin. And the logging data of the offset well is used to develop a logging data prediction model for the well under drilling based on the error Back Propagation Neural Network (BPNN). Then, a hybrid model is created by combining the Long Short-Term Memory Neural Network (LSTM) and the BPNN. Finally, the hybrid LSTM-BP model is trained and validated on the drilling and logging dataset, and the grid search algorithm is used to optimize the hyperparameters of the model. The hybrid LSTM-BP model is utilized to estimate formation pore pressure by inputting the drilling data of the well under drilling and longing data of the BPNN model predicted. The results indicate that the relative error of formation pore pressure prediction is less than 10%. This paper provides an accurate method to predict formation pore pressure in real-time, which is of great significance for maintaining the stability of the wellbore and ensuring the safety of drilling. INTRODUCTION Formation pore pressure is the pressure of the fluid in the formation, which is one of the most crucial fundamental data in the process of petroleum exploration and development, and a crucial foundation for the design of drilling plans and the analysis of wellbore stability (Azadpour et al., 2015; de Souza et al., 2021). However, when formation pore pressure is predicted incorrectly, it can lead to drilling risks including gas invasion, kick, and even blowout, which may result in enormous financial losses such as in the Deepwater Horizon tragedy. Therefore, accurate prediction of formation pore pressure in the process of drilling is conducive to ensuring drilling safety, improving rate of penetration (ROP), protecting the reservoir, and improving oil and gas recovery (Lei and Jie, 2004).