Wide-Area Control of Power Systems: Employing Data-Driven, Hierarchical Reinforcement Learning

Abstract

In the foreseeable future, the north American power transmission system is expected to evolve into the largest and most complex Internet of Things (IoT) technology, encompassing the use of massive volumes of data from phasor measurement units (PMUs) to make control decisions in the face of major disturbances without jeopardizing grid stability or the quality of service. In this modern grid, besides conventional generation, distributed energy resources (DERs) in the form of renewables, smart loads, and power electronic converters, will serve as active end points that not only consume, but can also generate, store, sense, compute, communicate, and react to datadriven commands. They will turn the grid into a large network of active nodes, introducing rapid, large, frequent, and random fluctuations in power flow, voltage, and frequency with a huge amount of model uncertainty while at the same time increasing our capability to learn and control them through a massive, data-driven IoT platform.