Voltage Balancing of a Combined SEPIC-Cuk for Bipolar DC Microgrid Using ANN-Based Model Predictive Control

DC microgrids are increasingly attractive for modern distributed energy systems owing to their efficiency, power quality, and compatibility with renewable generation. Among the various DC microgrid designs, the bipolar DC microgrid architecture holds unique advantages over the other architectures owing to its characteristic wiring. However, maintaining voltage balance and stability across the positive and negative DC buses in the bipolar dc microgrid (BDCMG) poses a significant challenge, especially under dynamic load and generation conditions. This paper presents a hybrid bipolar converter that combines a SEPIC and Cuk converter capable of feeding symmetric voltage into the bipolar DC bus. An Artificial Neural Network (ANN)-based on long prediction horizon Model Predictive Control (MPC) strategy is proposed for the control of the hybrid bipolar converter. The long prediction horizon makes the MPC have a stronger predictive control ability compared to a conventional MPC. The MPC serves as the expert providing data to train the proposed ANN. The ANN model is trained to predict system behavior and generate optimal control actions, reducing computational complexity while maintaining high accuracy in tracking reference voltages and currents. Extensive simulations are done in MATLAB/Simulink and prove the effectiveness of the proposed ANN-based model predictive control in ensuring precise voltage balancing of the converter which is essential for BDCMGs.