Reinforcement Learning based Approach for Virtual Inertia Control in Microgrids with Renewable Energy Sources

Abstract

The increasing penetration of distributed and renewable energy sources into power grids results in various control challenges. Among others is a problem of decreasing inertia, especially in the case of islanded microgrids with high penetration of a low-inertia power sources. One approach is based on a concept of additional virtual inertia control. In this paper we propose reinforcement learning based virtual inertia control with deep deterministic policy gradients based optimization algorithm. The proposed solution is demonstrated using standard topology of a microgrid and compares to H-infinity and optimally tuned PI controllers.

Topics & Concepts