Are we underestimating peak loads? Effects of temporal resolution, demand aggregation, and household characteristics in energy models

Accurate estimation of peak electricity loads is important for planning infrastructure and system costs. Prospective energy system models often emphasize total generation rather than instantaneous power draw, in part for computational tractability, and thus necessarily underestimate true peak loads with coarse timesteps and low-resolution representation of demand. As the structure of electricity demand changes (e.g., due to beneficial electrification), historical heuristics about acceptable timesteps for estimating peak loads might no longer hold. Here, we use a validated agent-based model of households in Aotearoa New Zealand to estimate peak errors under electrification and other conditions by generating 1-minute residential electricity demand profiles, then resampling with timesteps from 2 min to 24 h. Coarser timesteps increase peak load errors: 1-hour timesteps underestimate true peaks by 44% for one household (with implications for e.g., electric panel sizing), and 9% for 100 households (with implications for e.g., total dispatchable generation requirements). Error is sensitive to large insulation increases, solar PV, and electric vehicles. We recommend modelers should avoid coarse timesteps for applications like capacity expansion or battery sizing, where absolute peaks matter. We encourage modelers to carefully consider how changing system structure changes assumptions about acceptable simplifications, especially in safety-relevant contexts.