The importance of spatial resolution in large-scale, long-term planning models

Capacity expansion models are important tools for examining the evolution of the electric power sector. Embedded in these tools are many modeling choices with consequential impacts on computational burden and associated analysis. In this study, the spatial resolution of the national-scale Regional Energy Deployment System (ReEDS) model is adjusted to understand the implications of higher-fidelity modeling on energy system projections and model solve times. The default ReEDS regions capture the contiguous United States in 134 balancing areas, while the regions in the higher resolution version are defined by more than 3000 U.S. counties. Using both resolutions, a case study is conducted in the United States for the Texas Interconnection (ERCOT) and the Western Interconnection (WI) to explore how differences in spatial resolution impact model projections and to inform appropriate applications of high spatial resolution in a large-scale capacity expansion model. In both interconnections, the higher spatial resolution model achieves a lower-cost solution, attributed to the more detailed representation of variable renewable resources and transmission. Shifts in land-based wind capacity between the balancing-area-level model and the county-level model are more prominent than the changes in solar, in part because of the heterogeneity of wind resource across the United States and the stronger dependency of wind on transmission. Furthermore, at higher spatial resolution there is a locational shift in the installed capacity toward regions characterized by resource profiles that are better aligned to contribute to resource adequacy. Beyond the nuances in the results, running the high-fidelity ReEDS model introduces a significant computational burden, with an order of magnitude increase in the number of model regions leading to at least an order of magnitude increase in the runtime. Spatial flexibility can offer users and developers the opportunity to perform high-fidelity analysis, however the benefits of high-resolution modeling must be weighed against the availability of the necessary data and the scope of the research question.