Prediction of In-Situ Stresses by Using Machine Learning and Intelligent Optimization Algorithms

ABSTRACT In-situ stresses play an important role in affecting many geological processes such as hydraulic fracturing and CO2 storage, and a good understanding of the magnitude and direction of in-situ stresses is very important for deep energy exploitation. In this study, a machine learning model consisting of generative adversarial networks (GAN), particle swarm optimization (PSO) and support vector regression machine (SVRM) is proposed to predict the minimum in-situ horizontal principal stress (Shmin) from a series of existing experimental breakout data. First, the GAN and PSO are used to improve the quantity and quality of training data and to optimize the hyperparameters of the SVRM. Second, the enhanced training data are used to train the SVRM that predicts the Shmin based on the wellbore breakout geometries. In order to examine the reliability of this technique, the Shmin predicted from the proposed model is compared against the experimental data and it is found that the proposed model has a high accuracy with an average relative error of less than 10%. In addition, the proposed model requires only a few seconds to run on a laptop computer, thus providing a useful tool for accurate and efficient prediction of the Shmin. INTRODUCTION In-situ stresses play an important role in affecting many geological processes such as hydraulic fracturing and CO2 storage, and a good understanding of the magnitude and direction of in-situ stresses is very important for deep energy exploitation. During exploiting unconventional petroleum resources, drilling of a well and its stability depend on accurate measurement of in-situ stresses. Once the stress concentration at the borehole wall exceeds the rock strength, borehole failure occurs. In addition, in-situ stresses are closely related to earthquake faulting, volcanic eruptions and other geotechnical or geological processes. As a result, accurate measurement of in-situ stresses is of great significance.