Machine Learning to Discover, Characterize, and Produce Geothermal Energy

Abstract

Geothermal energy is an enormous, untapped, and always-on renewable resource that harnesses the natural heat of solid earth. The efficient discovery and deployment of this resource have various advantages for energy security and diversity. They allow us to improve domestic energy needs, provide flexible and affordable power to the nation's grid, and have economic consequences (e.g., jobs in the renewable energy sector). However, exploring, developing, and operating a geothermal resource is a significant challenge due to the high costs and risks associated with the drilling of wells, construction of the power plants, and maintenance of the facilities. Costs of a geothermal power plant are primarily early expenses, rather than fuel to keep them running. For example, cost of drilling a geothermal well can range anywhere between $5–20 million. Operating and maintenance costs range from $0.01–0.03 per kWh (i.e., $200 million to build a plant that produces 500 MW/hour). The initial cost for the field and power plant is around $2500 per installed kW in the United States, probably $3000–5000/kW for a small (<1 MW) power plant. Recently, substantial investments from the industry and government agencies have been made to reduce costs and risks associated with geothermal development. The focus is on applying advances in technologies and machine learning (ML) that address crucial exploration and operational challenges. ML methods, associated data analytics, and model diagnostics-based workflows allow us to reduce costs throughout the geothermal project life cycle (from resource exploration to power plant operations). This chapter discusses ML's role in discovering, characterizing, and producing geothermal energy within the United States. First, we provide details on the state-of-the-art algorithms and the data requirements to build accurate and reliable ML models. Next, we present some case studies and preliminary success stories of applying ML for the geothermal project life cycle. Building on the insights from these case studies, we identify challenge problems and high-priority research areas where ML can unlock the potential to accelerate geothermal resource discovery and energy production. Through the ML analysis presented in this chapter, we set the stage for broader applications of ML methods and tools, which researchers and stakeholders can utilize in the geothermal industry and institutions to overcome significant barriers in the geothermal project life cycle.