Distributed flexible resource regulation strategy for residential communities based on deep reinforcement learning

In an era characterized by the rapid proliferation of distributed flexible resources (DFRs), the development of customized energy management and regulation strategies has attracted significant interest from the field. The inherent geographical dispersion and unpredictability of these resources, however, pose substantial barriers to their effective and computationally tractable regulation. To address these impediments, this paper proposes a deep reinforcement learning-based distributed resource energy management strategy, taking into account the inherent physical and structural constraints of the distribution network. This proposed strategy is modelled as a sequential decision-making framework with a Markov decision process, informed by physical states and external information. In particular, targeting the community energy management system for critical public infrastructure and community holistic benefits maximization, the proposed approach proficiently adapts to fluctuations in resource variability and fluctuating market prices, ensuring intelligent regulation of distributed flexible resources. Simulation and empirical analysis demonstrate that the proposed deep reinforcement learning-based strategy can improve the economic benefits and decision-making efficiency of distributed flexible resource regulation while ensuring the security of distribution network power flow.