Proceedings of the PowerUp Conference

We design statistical methodology to project the impact of climate change on residential load at the level of substations. To achieve high-resolution load projections, we combine the outputs of downscaled global climate models (GCMs) with a statistical additive regression model. To this end we design a bespoke Generalized Additive Model (GAM) that links cooling/heating loads at given substation and hour of day to daily meteorological covariates. Our models are trained on the ResStock outputs for a realistic building mix and illustrated on the entire 1898 substations in the CATS synthetic grid. The results visualize feature relevance and aggregate climate impacts on California residential loads.