Fracture Pressure Prediction by Using the LSTM Neural Network Model

Abstract

ABSTRACT Fracture pressure plays a vital role in petroleum drilling and hydraulic fracturing. Currently, fracture pressure is mainly predicted through logging interpretation, but its calculation process is too complicated, and its accuracy needs to be further improved. Machine learning provides a new measure to address the above problems. Therefore, this paper aims to predict fracture pressure using the Long- and Short-Term Memory (LSTM) neural network method. The fracture pressure data set was generated by logging interpretation, then the nonlinear mapping relationship between logging parameters and fracture pressure was proposed using the LSTM model, and the mesh search method was used to optimize the hyperparameters of the LSTM model, so that the prediction of fracture pressure is realized. The results indicate that the in-situ stress state follows the normal faulting regime, i.e., vertical stress > maximum horizontal stress > minimum horizontal stress. Using the LSTM model with the optimal combination of hyperparameters, the prediction accuracy was significantly improved, and the root mean square error, mean absolute error, mean absolute percentage error, and coefficient of determination of the predicted fracture pressure were 0.304 MPa, 0.176 MPa, 0.209%, and 0.990, respectively. It is concluded that the LSTM model can effectively capture the variation trend of logging parameters with depth and the correlation information of logging parameters, which can realize the accurate prediction of fracture pressure. INTRODUCTION Fracture pressure is a very important fundamental engineering parameter that plays a critical role in oil drilling, completion and hydraulic fracturing (Chen et al. 2008; Fjar et al. 2008; Aadnoy and Looyeh 2011; Ma et al. 2017a). In drilling engineering, fracture pressure is the basis of drilling engineering design; if the fracture pressure is less than the effective wellbore pressure, the wellbore will experience tensile fracture (e.g., wellbore fracture or lost circulation), resulting in loss of drilling fluids. Predicting fracture pressure is therefore directly related to the safety of drilling operations. Incorrect or inaccurate prediction of fracture pressure can easily cause various drilling problems, such as lost circulation, induced well blowout, induced well collapse, and drill pipe sticking (Ma et al. 2017b).