Applications of artificial intelligence in geothermal resource exploration: A review

Abstract

Artificial intelligence (AI) has become increasingly important in geothermal exploration, significantly improving the efficiency of resource identification. This review examines current AI applications, focusing on the algorithms used, the challenges addressed, and the opportunities created. In addition, the review highlights the growth of machine learning applications in geothermal exploration over the past decade, demonstrating how AI has improved the analysis of subsurface data to identify potential resources. AI techniques such as neural networks, support vector machines, and decision trees are used to estimate subsurface temperatures, predict rock and fluid properties, and identify optimal drilling locations. In particular, neural networks are the most widely used technique, further contributing to improved exploration efficiency. However, the widespread adoption of AI in geothermal exploration is hindered by challenges, such as data accessibility, data quality, and the need for tailored data science training for industry professionals. Furthermore, the review emphasizes the importance of data engineering methodologies, data scaling, and standardization to enable the development of accurate and generalizable AI models for geothermal exploration. It is concluded that the integration of AI into geothermal exploration holds great promise for accelerating the development of geothermal energy resources. By effectively addressing key challenges and leveraging AI technologies, the geothermal industry can unlock cost-effective and sustainable power generation opportunities.

Highlights


  • Progress in the use of Artificial intelligence (AI) methodologies is presented in detail.

  • Geophysical data analysis is the most notable AI application.

  • Neural networks are the most-used AI technique across geothermal exploration groups.

  • Challenges and recommendations for future research using AI are provided.

  • Large-scale AI applications are reasonably novel in geothermal exploration.


1 INTRODUCTION

Geothermal energy is a renewable, sustainable, and low-emission energy source derived from the Earth's subsurface layers through natural heat sources, such as rock formation and radioactive decay. It is used for heating, cooling, and power generation due to its cost-effectiveness, stable supply, and high-capacity factors throughout the year. The development of power generation from hydrothermal reservoirs started in 1913 and has since expanded to include various technologies such as flash and dry steam plants for high-temperature resources and binary cycle technologies for medium-temperature resources. Global electricity generation from geothermal energy grew from 69.8 GW·h in 2011 to 95.3 GW·h in 2021, providing a significant share of electricity demand in countries, such as El Salvador, New Zealand, Kenya, and the Philippines, and more than 90% of heating demand in Iceland (International Renewable Energy Agency, 2023).

The development of geothermal energy takes place in successive stages, starting with surface surveys, followed by exploration drilling for resource realization. If the resource is proven, delineation drilling follows to confirm the extent of the reservoir's productivity and its development plan. Production drilling and power plant construction can commence once the resource has been confirmed and financial viability has been established. Preliminary studies, exploration, and delineation drilling require significant investment and involve high financial risk, which can hinder resource evaluation plans. For example, in a recent appraisal study for geothermal exploration in Indonesia, the World Bank estimated that the predevelopment program would cost approximately USD 30 million, assuming a minimum of three wells for greenfield development and at least two wells producing an acceptable level of steam for site exploration to provide satisfactory evidence or resource availability (The World Bank, 2012). Figure 1 illustrates the geothermal development project stages, the level of risk at each stage, and the associated percentage of cumulative project cost (Gehringer & Loksha, 2012; ©World Bank; The World Bank et al., 2012).

Details are in the caption following the image
Geothermal development project cost and risk profile throughout various project stages (Gehringer & Loksha, 2012; The World Bank et al., 2012; reproduced under the terms of the CC BY 3.0 IGO copyright licenses, ©World Bank).

Many countries are exploring hidden or blind geothermal resources (hydrothermal resources without surface manifestations), which require detailed knowledge of subsurface features (including hydrological, geophysical, geological, geomechanical, geochemical, and thermal characteristics) to assess their commercial potential (Pandey et al., 2018). Traditional methods of subsurface feature analysis rely heavily on expert knowledge for resource evaluation and reserve estimation, leading to uncertainties in the discovery of hidden geothermal resources. Advances in data-driven models have led to the use of artificial intelligence (AI) to replace traditional expert-based and statistical methods, where AI can uncover hidden patterns and develop predictive models from large multivariate datasets, thus enhancing exploration outcomes by reducing uncertainty and improving prediction accuracy. With the rapid increase in the creation of data repositories for the preservation, processing, and management of subsurface data, data-driven models offer an efficient and cost-effective approach to identifying key features of hidden geothermal resources (He et al., 2019). This supports resource evaluation, problem solving, and decision-making while reducing predevelopment costs in the geothermal industry.