Triple-optimized droop control for DC microgrids: A machine learning approach for minimizing cost, loss, and enhancing flexibility

This paper presents a triple-optimized machine-learning-based droop control strategy that can reduce the production expense (PE) and power delivery losses (PDL) at the same time and increase the operational flexibility of direct current microgrids (DCMGs). Traditional approaches such as hybrid droop techniques have partial PE and PDL improvements but do not employ a systematic way of increasing flexibility and often use constant weighting coefficients (WCs). In order to address these limitations, the suggested approach in this study incorporates artificial intelligence-based predictive models, which help to estimate PE, PDL, and flexibility indices. A multi-objective optimization algorithm based on a gradient descent technique is then used to optimally and simultaneously enhance all three performance measures under different operating conditions. The effectiveness of the proposed strategy is confirmed through various simulations in the Matlab/Simulink environment under different scenarios, including sudden changes in loads. The results show that the proposed method reduces PE by approximately 4%, 20%, and 21% compared to the classical, L D , and F D methods, respectively, while improving PDL and flexibility simultaneously. However, due to the multi-objective nature of the optimization, the PE is slightly higher (approximately 8%) compared to the I C D method, reflecting the inherent trade-off among competing objectives. Overall, the proposed framework provides a more balanced and flexible performance compared to existing approaches.