Development of Efficient Empirical Models for the Prediction of Oil Well Fracture Pressure Gradient

Abstract

Evaluation of fracture pressure gradient during oil well drilling has been best achieved in the past using the leak-off test. This study however, utilized artificial intelligence techniques involving Hunter Point-Artificial Neural Network (HP-ANN), Multivariate Regression (MVR), and Adaptive Neuro-Fuzzy Inference System (ANFIS) to develop a prediction model for fracture pressure gradient based on the input parameters of pore pressure, vertical depth, fracture pressure, and overburden pressure. The dataset used for training the models were extracted from the works of Akinbinu 2010 and Udo et al. 2020. The three models' prediction performance was compared with existing literature models using RSME, MAE, and R2 error analysis indicators. The HP-ANN model was found to have the highest prediction accuracy for oil well drilling fracture pressure gradient. Using the optimum HP-ANN model weights and the biases, an empirical mathematical equation was extracted for the prediction of the Fracture pressure gradient. Along these lines, the created HP-ANN models can be utilized to predict the Fracture pressure inclination of an oil well for practical purposes.