As variable renewable energy resources become a larger part of the generation mix in the United States (U.S.), so does the potential impact of prolonged periods of low wind and solar generation, known as variable renewable energy (VRE) droughts. In a future decarbonized or low-carbon grid, naturally occurring VRE droughts need to be evaluated for their potential impact on grid reliability. This study is the first of its kind to examine the impacts of compound VRE energy droughts in the Western U.S. across a range of climate change and future infrastructure scenarios. We find that compound VRE drought severity will increase significantly in the future, primarily due to the dramatic increase in wind and solar generation needed to meet decarbonization goals. Climate change is expected to increase the variability of energy drought severity, which has implications for sizing energy storage necessary for mitigating drought events. We also examine the spatial patterns of compound VRE drought events that effect multiple regions of the grid simultaneously. These co-occurring events have distinct spatial patterns depending on the season. We observed overall fewer connected events in the future with the combined effect of climate change and infrastructure growth, although in the fall we observe a climate change-induced shift toward events which impact more regions simultaneously.

Most hydropower utilities rely on flow forecasts to manage the water resources of their reservoir systems and to help marketers and schedulers make efficient use of power generating resources. Flow forecast providers and dam operators typically assess the value of flow forecasts by assessing the skill of the forecasts in a verification exercise. Although there are many flow forecasting approaches available—from physics-based approaches associated with statistical pre and post processors and data assimilation, to emerging machine-learning based approaches—there is little consensus on how to choose the best forecast product. Nor are there established methods for translating forecast skill—a summary statistic amalgamating multiple types of errors —to forecast value (benefits or avoided cost) as perceived by a marketer or scheduler. In this work we develop such an approach by combining a water resources management model with a power grid model. Flow forecasts are developed at 85 locations for a varying range of skills, from perfect, to persistent and in-between. Using reservoir and power grid simulations over the Western U.S., we propagate flow forecasts through the power grid model, mapping flow forecast skill to regional hydropower revenues, production costs and carbon emissions. We develop a deeper understanding of the influence of regional and seasonal differences in markets and hydrologic dynamics on forecast value. We discuss future research directions to integrate hydrologic forecasts into decision-making at the utility and wider system scale.