Optimization of Power Flow in Hybrid Microgrids Using AI-Based Algorithms

Abstract

Increased renewable energy applications in microgrids imply that there is currently less predictability in power generation, implying that there is a necessity of sophisticated methods to streamline the operations of the microgrid so that it operates effectively and steadily. In this work, a hybrid Artificial Intelligence (AI) framework is proposed to enhance the way power is transferred in AC and DC microgrids with energy storage systems (ESS), distributed energy resources (DERs) and integrated with the normal utility grid. The proposed solution comprised of DDPG Deep Reinforcement Learning and Genetic Algorithms address both real-time and global optimization issues. The DRL agent determines the optimal strategies to exploit RES, ESS and the grid and the GA enhances the initial parameters and critical elements of the model of the agent to achieve optimal performance. The framework is studied under different conditions and load and generation patterns in two simulation environments, one in MATLAB/Simulink and the other in Python during 24 hours. Compared to the outcomes, this DRL-GA approach is a far better solution than the Mixed Integer Linear Programming (MILP) and Particle Swarm Optimization (PSO), leading to an operation cost reduction, enhancing voltage stability as well as increasing renewables utilization. The findings provide a flexible and smart control strategy that will be efficient in microgrid energy management.