ANN-Based Model Predictive Control for Hybrid Energy Storage Systems in DC Microgrid with Constant Power Loads

This study presents, an innovative approach to address computational challenges associated with conventional model predictive control (MPC) termed as artificial neural network (ANN) based MPC for DC microgrids. In the proposed technique, MPC is employed as an expert, providing data to train the ANN. Once finely tuned, the ANN is used to control the DC-DC converters eliminating the need for extensive mathematical computations. At the same time, the increased integration of DC-DC converters in DC microgrids introduces challenges, particularly in addressing constant power loads (CPL) instability. On this basis, the scope of this study extends to DC microgrids, specifically focusing on a hybrid energy storage system (HESS) designed to balance power fluctuations in DC microgrid with CPL. The proposed ANN-MPC is integrated with decentralized power sharing strategy, composed of virtual impedance droop control for fuel cell (FC) and battery. Therein, FC supports steady state loads, while the battery handles rapid load variations. Initially, the proposed controller is designed for single energy storge system and then coordinated with HESS. Simulation results on MATLAB/Simulink demonstrate the effective performance of the proposed controller compared to conventional MPC, particularly in achieving smoother transient behavior.