Enhancing Hydropower Resilience and Grid Stability with Interpretable and Customizable AI Amid Climate Change

Climate-driven water cycle changes have led to a global hydropower decline, exacerbating power shortages and grid risks. Here we developed HydroTrace, an AI model featuring a customizable attention mechanism rooted in Earth system processes. It enables hydropower stakeholders to identify the climate and environmental factors that have the most significant impact on their operations. HydroTrace outperforms conventional methods in the Third Pole, a region heavily reliant on hydropower, sensitive to climate changes, and lacking in data. It reveals that glacier-related runoff significantly affects hydropower generation. Projections up to 2100 predict increased spill periods (when runoff exceeds a hydropower station’s capacity) by 13-21 days and decreased low-flow production by 14.0-19.3%. This leads to an overall reduction in generation by 0.4-4.4% at the facility being studied. With its precision and adaptability, HydroTrace serves as a vital tool for the development of climate adaptation strategies and the formulation of energy sector policies.