基于多视角迁移学习的风场内机群划分及等值风场参数综合优化

To ensure the multi-scenario applicability of equivalent wind farm(WF) model, this paper proposes a wind turbines(WTs) clustering method based on multi-view transfer learning, and construct an optimization model of equivalent WF parameters, and solve it by high-dimensional multi-objective evolutionary optimization algorithm. Firstly, multi-scale entropy(MSE) of active power, reactive power, voltage and current of WT is used as the clustering indicator, and the applicability of multi-view indicators in clustering WTs is analyzed. To improve the multi-scenario applicability of the equivalent WF model, and taking into account the multi-view characteristics of the clustering indicator, multi-view fuzzy C means(MV-FCM) clustering and transfer learning are combined. A new clustering algorithm, multi-view transfer fuzzy C means(MVT-FCM) clustering algorithm is proposed for clustering WTs. Next, considering the equivalence precision of active power, reactive power, voltage and current, the equivalent WF parameter calculation is transformed into high-dimensional multi-objective optimization problem, and the knee point-driven evolutionary algorithm(KnEA) is adopted to solve it. Finally, a case study of 16 WTs in WF and an actual WF in a certain area is carried out. The results verify the accuracy and multi-scenario applicability of the WF equivalent model.