基于Q学习粒子群算法的海上风电场电气系统拓扑优化

针对多变电站海上风电场的电气系统拓扑优化,现有的方法一般根据预先确定的变电站个数将整体海上风电场划分为几个固定的子区域,然后分别进行独立的电缆连接布局优化,最终聚合得到整体方案.然而,采用固定的划分策略很难得到全局最优方案.因此,考虑多海上变电站选址、电缆选型、功率损耗等因素,以最小化成本为目标,建立多变电站海上风电场的电缆连接布局优化模型,并提出一种基于Voronoi自适应分区的Q学习粒子群算法进行求解.所提出的算法以Q学习粒子群算法为核心,设计一种基于Voronoi图的自适应分区策略实现自适应分区,并结合相应的编解码策略实现不同分区的电缆连接.最后,通过算例分析证明所提出模型以及算法的有效性.

Optimization of Electrical System Topology for Offshore Wind Farm Based on Q-learning Particle Swarm Optimization Algorithm

For the electrical system topology optimization of multi-substation offshore wind farms, the current methods are to divide the overall offshore wind farm into several fixed sub-areas according to the predefined number of substations, then independently optimize the cable connection layout in each sub-area and eventually combine all of them as the overall scheme. However, it is often hard to obtain the global optimal scheme due to the fixed partition strategy. Therefore, this paper designs a cable connection layout model for multi-substation offshore wind farms and proposes a Q-learning particle swarm optimization algorithm based on Voronoi adaptive partition, which aims to minimize the total cost considering substation locations, cable type selection, and power loss. Taking the Q-learning particle swarm optimization algorithm as the key, the proposed method designs an adaptive partition strategy based on the Voronoi diagram, which can realize the cable connection in different partitions with coding and decoding strategies. Finally, the case analysis proves the effectiveness of the proposed model and algorithm.