Reconfiguration of active medium- and low-voltage distribution networks based on multi-agent systems and mutual support capacity

Abstract

Reducing power losses and preventing voltage instability remain key concerns in electrical distribution networks (DNs). These challenges are often addressed through reconfiguration of the network and the integration of distributed generation (DG) resources. In this study, a multi-agent system is proposed to reconfigure DNs and to determine the appropriate capacity of DG units to minimize both power losses and voltage deviations. Although medium-voltage (MV) and low-voltage (LV) networks function as separate systems with different operating requirements, their decisions are interdependent, with each influencing the overall objective function and decision space of the other. The proposed approach accounts for this interaction by considering both MV and LV networks simultaneously, while also exploiting the potential of mutual voltage support between the two levels. To capture realistic operating conditions, uncertainties in renewable generation and load demand are included in the analysis. The method is applied to a test system comprising an MV network connected to three LV feeders, and optimization is performed using a genetic algorithm. Simulations were conducted in MATLAB with the MATPOWER 8.0 toolbox. To validate the proposed method, it is compared under several scenarios. The simulation results indicate that the proposed architecture outperforms the existing architectures. Specifically, in the MV DNs, the proposed architecture reduces active power losses by up to 44.70% and 25.80% compared to the initial configuration without DG and the reconfigured network without DG, respectively. Moreover, a loss reduction of 27.99% is achieved compared to DG‑based reconfiguration using a centralized approach in the MV network. For the integrated MV–LV network, the proposed method further reduces total power losses by 5.57% and 5.06% compared to DG‑based reconfiguration using centralized and decentralized MAS approaches, respectively. In addition, voltage deviation is significantly improved, and the voltage profiles remain nearly uniform and stable during different operating hours.