Adaptive Voltage and Frequency Regulation for Secondary Control via Reinforcement Learning for Islanded Microgrids

This paper demonstrates an innovative method to enhance the secondary control of microgrids in islanded conditions through the implementation of a Reinforcement Learning (RL) Controller based on Proportional-Integral (PI) principles. The aim is to adjust microgrid's voltage and frequency. The intelligent microgrid, incorporating diverse energy sources, represents the future of power grids. Ensuring its dependable operation requires sophisticated communication and clever data handling. Within this structure, Distributed Generations (DGs) utilize localized and worldwide control loops. for voltage and frequency regulation. A novel distributed secondary control scheme, employing a Reinforce-ment Learning (RL)-based Proportional-Integral (PI) controller, dynamically adjusts its parameters for optimal performance. The system consists of two Distributed Generators (DGs) linked to each other. The performance of the system is investigated under the islanded condition. The RL-based PI controller which uses the TD3 algorithm for training the agent outperforms the Conventional PI controller, demonstrating faster frequency and superior voltage regulation upon microgrids becoming islanded.