Data-driven, Reduced-order Model Representation of Load and Generation Shedding Schemes

Data-driven, reduced-order models of power system frequency control are becoming important analytics tools to obtain fast assessments of dynamic security in power systems. However, these models are commonly associated with low precision in terms of representing more complex phenomena, such as intentional islanding or load and generation shedding schemes. In this way, its applicability to predict or study extreme frequency events has been questioned. This work presents a data-driven algorithm to represent load/generation shedding events using a reduced-order model of frequency dynamics. The applicability of the model is examined by considering an actual case of a cascade load/generation shedding event in the Chilean power system.