Zonotope and Reinforcement Learning-based Aggregator Operation of Energy Community Resources with Water Pump

Flexible distributed resources within the energy community at the distribution network side have considerable potential for grid-interaction. However, the aggregation and centralized scheduling of these resources confront inevitable challenges in terms of computational complexity and decision-making efficiency. This paper proposes an efficient and computational tractable aggregation and scheduling method for flexible distributed resources based on Zonotope and deep reinforcement learning (DRL) technology. Notably, this study addresses the con-current scheduling dilemma involving large-scale flexible energy devices and numerous small-scale distributed resource clusters, which would capture the characteristics of group intelligent decision-making that evolves as energy status changes. Through typical numerical results with simulation experiments, this work demonstrates and showcases a considerable enhancement in both economic benefits and decision-making efficiency for the aggregator of energy community using the intelligent scheduling framework.