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

Wind Turbines Clustering in Wind Farm Based on Multi-view Transfer Learning and Synthetic Optimization of Parameters in Equivalent Wind Farm

为提高等值风场模型精度和多场景适用性,提出一种基于多视角迁移学习的风场内机群划分方法,构建了等值风场参数综合优化模型,并采用高维多目标进化优化算法对模型进行求解。首先,将风机出口有功功率、无功功率、电压和电流的多尺度熵(multi-scale entropy,MSE)作为机群划分指标,并分析了多视角指标在机群划分中的适用性;其次,为减少机群划分次数,以提高等值风场的多场景适用性,将多视角模糊C均值(multi-view fuzzy C means,MV-FCM)聚类与迁移学习有机结合,提出一种新的聚类算法——多视角迁移模糊C均值(multi-viewtransferfuzzyCmeans,MVT-FCM)算法,用于机群划分;接下来,为进一步提高等值风场仿真精度,综合考虑有功功率、无功功率、电压和电流的等值准确性,将等值风场参数计算转化为高维多目标优化问题,并采用膝点驱动的进化算法(kneepoint-driven evolutionary algorithm,Kn EA)进行求解;最后,对含16风机风场、某地区实际风场以及某风电汇集区域分别进行算例分析,结果验证了等值风场模型的精确性和多场景适用性。

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.