Multi-Agent Energy Allocation Optimization in Local Energy Communities: A Comparative Study

Abstract

The growing diversity of energy demands, combined with the integration of renewable and distributed energy systems, calls for more adaptive energy management strategies. This study presents a comparative simulation-based analysis of three energy management approaches applied to a local energy community: A Multi-Agent System (MAS), a centralized rule-based method, and a game-theory-based optimization using the Alternating Direction Method of Multipliers (ADMM). The MAS approach models buildings, Electric Vehicles (EVs), and energy storage systems as autonomous agents that dynamically allocate energy based on user preferences. Simulations were conducted using the Multi-Agent Simulation Environment (MESA) framework in Python, with a focus on optimizing energy allocation while minimizing costs and ensuring user comfort. This decentralized approach enables each agent to make local decisions while collectively achieving system-wide objectives. The comparison uses the same use case and dataset across all three methods, ensuring methodological consistency and strengthening the reliability of the performance evaluation. The results demonstrate that the MAS approach achieves lower overall energy costs compared to the rule-based method and ADMM in scenarios prioritizing balanced energy distribution and self-sufficiency, where 'balanced' refers to a scenario that equally weighs user comfort, cost, and local renewable usage objectives. The MAS achieves a total community cost of C359.72 per day in the balanced scenario, compared to C395.54 per day for the rule-based approach and C375.94 per day for the ADMM method, representing cost savings of 9.1 and 4.3%, respectively.