Predicting Formation Breakdown Pressures for Arbitrary Wellbore Orientation Using Machine Learning Techniques

One of the major operational challenges facing hydraulic fracturing operations is the accurate estimation of the associated formation Breakdown Pressure. Underestimating this value can lead to frac-job cancellations, and overestimating it can result in huge expenditure loss. The complexity of the problem stems from the fact that there are several factors affecting this value, some of which are fluid-related parameters, while others are mechanical-related parameters. Conventional techniques for estimating this value rely on empirical formulas which try to approximate (or suppress) the effects of these different parameters in one or several equations. This does not necessary lead to accurate results, especially for highly heterogenous reservoirs. In this work, we will develop a machine learning approach, using Neural Networks, for computing breakdown pressure values for an arbitrary wellbore orientation. The novelty of the work presented in this paper lies in the way the training data set is prepared for the constructed Neural Network. Due to the scarcity of the training data, we will utilize the computational framework, presented in (Almani et al., 2021 (1,2)), for generating the training data-set for our Neural Network. Once the Neural Network is trained, it will be used to predict formation breakdown pressures for different wellbore orientations.