Mitigating Climate Risk to China's Renewable Energy Generation by Power Plants Redeployment Using a Machine-Learning Model

Redeploying plants may mitigate climate risk and enhance renewable power generation. However, designing deployment strategies is complicated by the lack of plant-level response between power generation and climate variables with the constraint of economic and social factors. Here, we develop three random-forest (RF) response models that accurately capture the nonlinear relationship between renewable energy generation (hydro, solar, and wind power) and climate parameters at the plant level, using a 17-year historical data set. These RF models enable projections of renewable energy generation from both existing and newly built power plants under the Representative Concentration Pathways and the Shared Social-Economic Pathways (RCP-SSP) scenarios, as well as deployment strategies. Our analysis reveals that renewable energy generation from existing plants is projected to decrease significantly by 6%–8% (57–72 TWh) in 2045–2060 compared to the period 2002–2017. The impact of climate change on renewable energy generation varies spatially, suggesting optimizing the deployment of newly built power plants could mitigate adverse effects. Compared to the strategy maintaining the original deployment, national renewable energy generation can be increased by 24%–28% through optimized deployment tailored to future climate. The optimized deployment can lead to synergistic reductions in carbon emissions by 25%–28% and air pollutants by 42%–97%. These findings underscore the significance of considering plant-level heterogeneity and climate risk in the strategic deployment of renewable power systems.