Improving Current Sharing and Voltage Regulation for DC Microgrids: A Decentralized Demand Response Approach

Abstract

Droop control is a widely applied method to improve the current sharing and voltage regulation of each distributed generator in a DC microgrid. However, the droop coefficients are usually fixed and assigned based on the designed capacity of the distributed generators, leading to fluctuations in the voltage and current at the DC buses. This paper introduces a novel droop control based on a decentralized demand response approach with DC boost converter, to derive the optimal droop coefficients. In the proposed model, the switch states (i.e., closed, and open) of the DC boost converter at each distributed generator are utilized to employ the decentralized demand response for dealing with the uncertainties in the output power of distributed generators and the resistive loads. A multi-objective optimization approach based on reinforcement learning and interactive fuzzy programming is proposed to regulate the system voltage, minimize the total power loss, and improve the current sharing error among the distributed generators. First, the Pareto solutions are obtained using the deep-Q network-based reinforcement learning method to handle the uncertainties. Then, the best compromise solution is selected based on the interactive fuzzy method with a modified aggregation membership function to consider both dominant and compromise values for an optimal solution. The proposed approach is tested with a parallel-connected DC modified 30-bus system. Further, its performance is also validated by comparing it to the centralized demand response-based droop control and the classical stochastic method.