Investigating the Efficiency of Neural Networks in Predicting Breakdown Pressures for Tight Gas Formations

ABSTRACT: Determining accurate formation breakdown pressures during hydraulic fracturing operations is one of the major operational challenges for tight sandstone formations in compressional in-situ stress regimes. An accurate estimation of the corresponding breakdown pressure has a direct impact on the cost of the associated hydraulic fracturing job in the sense that underestimation can lead to operational failures and frac-job cancellations and overestimation results in inappropriate selection of the well completion type, and, in turn, huge expenditure loss. To address these challenges, several approaches for predicting formation breakdown pressure exist in the literature. Lots of these approaches rely on empirical formulas approximating the different factors affecting the breakdown pressure such as tectonic stresses, wellbore and pore pressures, and poro-elastic stresses. In our approach, we will investigate the efficiency of Artificial Neural Networks (ANNs) in predicting formation breakdown pressures. The novelty of our approach comes from the fact that the training data set of the neural network is obtained through a unique hybrid analytical and computational approach for computing breakdown pressures, which is rigorously calibrated against measured Breakdown Pressures in the field. Once trained with a sufficient calibrated data set, the ANN can be used to predict formation breakdown pressures given arbitrary input data sets. This is still a work in progress, but the preliminary obtained results reveal a great potential in the use of ANNs in obtaining reliable models for predicting formation Breakdown Pressures.