Voltage Regulation in Distribution Network using Bi-Level Multi-Agent Optimal Energy Management

In this paper, a bi-level optimal energy management with a techno-economic objective of regulating the voltage and mitigating economic losses in a distribution network (DN) with 100% rooftop photovoltaic (PV) penetration is presented. A deep reinforcement learning (DRL) agent is used at the network level to achieve the techno-economic objectives of the distribution system operator (DSO). Within DN, each smart home with energy management system (HEMS) uses DRL for effective energy management when responding to the economic signal broadcasted by the DSO agent. Further two different methods of generating economic signals, i.e., dynamic price generation at different levels of HEMS penetration are studied. The results indicate that the proposed bi-level optimal energy management is effective in mitigating the undervoltage scenario using demand response at all levels of HEMS penetration. However, the effectiveness of the proposed system in regulating the overvoltage in the DN is observed only at 100% HEMS penetration.