Machine learning in subsurface geothermal energy: Two decades in review

• Published in: Geothermics Vol. 102; p. 102401
• Main Authors: Okoroafor, Esuru Rita, Smith, Connor M., Ochie, Karen Ifeoma, Nwosu, Chinedu Joseph, Gudmundsdottir, Halldora, (Jabs) Aljubran, Mohammad
• Format: Journal Article
• Language: English
• Published: Oxford Elsevier Ltd 01.06.2022; Elsevier Science Ltd
• Subjects: Algorithms; Artificial intelligence; Decision making; Deep learning; Drilling; Drilling machines (tools); Exploration; Geothermal energy; Geothermal power; Geothermal resources; Learning algorithms; Machine learning; Problem solving; Reservoir engineering; Reservoirs; Resource development; Reviews; Seismicity; Statistical learning; Deep learning; Statistical learning; Artificial intelligence; Geothermal energy; Machine learning
• ISSN: 0375-6505; 1879-3576
• DOI: 10.1016/j.geothermics.2022.102401

•The application of machine learning in the geothermal industry is growing.•Years 2020 and 2021 saw a sharp increase in machine learning applications, especially the use of deep learning algorithms.•Reservoir characterization was the research area with the most publications.•Exploration, drilling and research on seismicity would benefit from more applications of machine learning. This paper reviews the trends in applying machine learning to subsurface geothermal resource development. The review is focused on the machine learning applications over the past two decades (from 2002 to 2021) to determine which machine learning algorithms are being used. In addition, the review seeks to determine what types of problems are being addressed with machine learning and how machine learning is aiding decision-making and problem-solving for subsurface aspects of the geothermal industry. The study shows that there has been a steady increase in the application of machine learning in the geothermal industry over the past 20 years, with an exponential increase in machine learning applications from 2018 to 2021. Several research areas associated with geothermal resource development were reviewed, including exploration, drilling, reservoir characterization, seismicity, petrophysics, reservoir engineering, and production and injection engineering. The study reveals that the field of reservoir characterization had the most significant applications of machine learning in the geothermal industry. Though machine learning has been applied across all the geothermal research areas we investigated, this study shows that there are still opportunities to improve and expand the adoption of machine learning in exploration, drilling, and seismicity. The main challenges that would need to be addressed are ensuring researchers have access to data, curating the data to be suitable for machine learning, and training geothermal industry students and professionals on artificial intelligence related to the energy sector.