Reinforcement-Learning Control for Coordinated Voltage Regulation in Distribution Systems Using a Power Hardware-in-the-Loop Platform

The variable nature of Distributed Energy Resources (DERs), such as solar Photovoltaic (PV) arrays and Wind Turbine Generators (WTGs), create several challenges for utilities and system operators related to maintaining voltage and frequency stability. Power grids that incorporate a variety of distributed energy generation must be capable of maintaining system voltage in the allowable range against potential disturbances caused by load variations and variable energy generation. Advances in grid support functionalities enable WTGs to provide reactive power in order to provide voltage regulation. New grid standards are requiring DERs to provide voltage regulation across distribution networks. Volt-Var Curve (VVC) control is an autonomous grid-support function that provides voltage regulation based on the relationship between voltage and reactive power. However, VVC only provides voltage regulation at the point of DER interconnection, and regulating actions follow a pre-defined rule with little flexibility. Therefore, there is a need for data-driven control methods, such as those based on Reinforcement Learning (RL), which can leverage system-wide measured data, and provide a customized and adaptive solution for voltage regulation across the entire system. Herein, this paper investigates the ability of a WTG to provide voltage regulation in a distribution system by comparing the performance of a traditional VVC and an RL-based data-driven control in a Power Hardware-in-the-Loop (PHIL) platform. The data-driven model training involves a MATLAB/Simulink simulation of a distribution system consisting of three WTGs, one solar PV array, and one variable load. Based on the obtained simulation results, the voltage regulation controls are evaluated in the PHIL platform that leverages a physical wind turbine emulator and a commercial PV inverter.