Optimization of Collaborative Planning for Source-Grid-Load-Storage in Green Park Based on Enhanced DDPG Method

Driven by the "dual-carbon" goals, the efficient integration of green power has gradually become a core proposition in the construction of new power systems. The source-grid-load-storage integrated collaborative management of green power industrial parks is one of the important approaches to achieve green power integration. With diverse ways to obtain green power, how to construct a reasonable green power supply strategy under the premise of ensuring life cycle cost optimization is a critical issue faced by current green power parks. In response, this paper proposes an optimal configuration method based on an improved Deep Deterministic Policy Gradient (DDPG) deep reinforcement learning algorithm. Firstly, an optimal model of collaborative planning for Source-Grid-Load-Storage in green park is constructed. Secondly, based on the architecture of the DDPG algorithm, the optimization problem is modeled as a Markov process. Finally, the effectiveness of proposed method is verified by the improved IEEE 33-node test system. By introducing the scene memory mechanism and combining with the delayed update structure, this proposed DDPG method effectively solves the problems of low sample efficiency and overestimation of Q value in the traditional actor critical algorithm in high dynamic environment and high-dimensional action space.