Geothermal, Oil and Gas Well Subsurface Temperature Prediction Employing Machine Learning

Abstract

Geothermal energy is getting more and more attention these days due to its nature of being a clean source of renewable energy sources provider with a zero-carbon footprint, free and available year-round. Assessment of this geothermal system is often neglected in practice during the exploration phase of energy extraction in geothermal development projects. The artificial intelligence expertise encompasses a vast number of deep learning and machine learning techniques ranging from linear regression, logistic regression, convolutional neural networks, genetic algorithms, reinforcement learning experiments, generative adversarial networks, etc. On amalgamating these available machine learning techniques with the geothermal concepts like power conversion, geology, thermodynamics, geophysics, electricity generation, heat flux, etc. are capable of improving the geothermal sector holistically by Subterranean resource characterization, Discrete micro-seismic event detection, and classification, drill fault deception, and prediction, predicting subsurface temperature and geothermal gradient and more. This study targets the analysis and prediction of the subsurface temperature of geothermal oil and gas wells. The implemented Support Vector Regression (RMSE = 4.75) shows better results when compared with the counter model, XGBoost (RMSE = 5.16).