基于SPP拓扑的海上风电场直流集电系统布局优化

With the rapid development of global offshore wind power, the scale and capacity of offshore wind farms (OWF) are continuously expanding, making it crucial to enhance the overall economic efficiency of OWFs. However, previous studies on DC collector systems of OWFs mainly focus on the DC series-parallel (SP) topology, which escalates the overall costs. To optimize the collector system layout, this paper proposes a novel hierarchical reinforcement learning (HRL) based framework for improving the overall economic efficiency by leveraging an advanced DC series-parallel-parallel (SPP) topology. In the proposed framework, a hierarchical open-loop multiple travelling salesman problem (HOMTSP) is utilized to model the SPP topology, decomposing the collector system layout optimization (CSLO) problem into sub-problems for resolution. Subsequently, a hierarchical double Q-learning (DQL) is employed to solve these sub-problems, with a topology-guided mechanism to refine the routing results and correct crossed cables by incorporating topological characteristics. Furthermore, this study acquires the GIS data and the connection scheme of wind turbines in a real OWF for the case study. Numerical results show the SPP-based framework significantly improves the economic efficiency compared to the DC SP topology and the AC double-sided ring topology.