Advanced MARL-Based Control Of Decentralized Battery Energy Storage Systems

Solving power storage management problems in microgrids and smart grids to support applications such as vehicle-to-grid has become a priority in the drive to clean energy. Inaccurate synchronization of the batteries’ charge-discharge levels is a known problem. Susceptibility to resulting control and instability issues is directly proportional to environmental and operational influences, such as dynamic environments with a high probability of load branch and storage switching, batteries’ heterogeneity, the DC impact of infrastructure, unmonitored faults, ineffective organization of energy management, and temperature disparities. This paper suggests and practically tests in real-time a nominee solution through a developed decentralized MARL-based controller. Accordingly, enhanced performance and accuracy of the insertions/removals (plug-and-play) of the batteries, and improved steadiness of output voltage are achieved with the existence of real operation influences in a DC self-governing microgrid. Specifically, multiagent “neighbor-to-neighbor” transfer of information was exploited to balance the level of involvement in consumer consumption of the battery energy storage system concerning its neighbors. Furthermore, real-time Energy-flow organizing was introduced to implement a plug-and-play policy under qualified compensation for the DC influence on the control. Hence, local energy flow consumption/loss was organized and monitored. Thus, regional power flow efficiency can be evaluated. Real-time Simulation and practical studies in critical operations and realistic conditions demonstrated a Key observed outperformance including an average, heightening of plug-and-play performance (0.66-13.366%), improved output voltage steadiness (2.637-3.24%), better stability and smoothness of battery’s voltage (2.9739-3.8462%), developed steadiness of load (6.666-37.091%), power consumption saving (2.94%), power flow balance enrichment (6.468 %), and improvement of power flow efficiency (2.626%), can be reached from the experimentations in the case when the developed strategy is employed.