Enhanced Representations of Thermal Generator Outage Risk in Capacity Expansion Models

Power system planning models have traditionally used installed capacity heuristics such as planning reserve margins to enforce resource adequacy constraints in least-cost capacity expansion problems. While these approaches are simple to implement and solve, they necessarily discard details about the non-linear interactions between and within different classes of resources, information that could be used by a model to better inform planning decisions.This paper describes two alternative appproaches that allow for direct consideration of probabilistic shortfall risk criteria in the optimization process, with tradeoffs between representational fidelity and computational performance. The accuracy and consistency of each method is compared empirically against a more traditional reserve margin-based approach. Unlike that method, which requires prescribing an installed capacity surplus that may not necessarily correspond with desired reliability outcomes, the two alternative methods endogenously consider thermal unit size and outage probabililty, yielding optimal reserve margin information as a model output, not an estimated input.