With the widespread integration of distributed generation into distribution network, the dispatching methods of distribution networks will undergo fundamental changes. Microgrids, as a way to absorb renewable sources, play an increasingly important role in new power systems. The coordination between distribution network and microgrid cluster helps promote the absorption of distributed resources, but traditional optimization methods struggle to quickly solve the complex problems of distribution network and microgrid cluster. Therefore, this paper proposes a multi-level collaborative dispatching method for distribution network and microgrid cluster based on safe deep reinforcement learning. First, considering the collaborative relationship between distribution network and microgrids, a multi-level dispatching model for distribution network and microgrid cluster is constructed. Then, the collaborative dispatching of distribution network and microgrid cluster is expressed as a Markov decision process. Addressing the constraint overshooting problem encountered when agent algorithms solve practical application problems, a multi-agent soft actor-critic with Lagrangian method algorithm is adopted. Finally, a case study is conducted based on the improved IEEE 33-node system to verify the effectiveness of the proposed model and algorithm.
Zhu et al. (Thu,) studied this question.