基于投票策略的智能配电网柔性负荷聚类分析方法

A cluster analysis method for flexible loads in intelligentdistribution networks based on voting strategy

面向智能配电网台区负荷特性差异大,以及传统聚类算法的聚类效果、聚类精度以及鲁棒性难以平衡的问题,结合K均值(K-means)算法和基于灰狼优化的模糊C均值(fuzzy C-means clustering algorithm based on Grey Wolf optimization, GWO-FCM)算法等不同算法的聚类稳定性和聚类效应,提出一种基于投票策略的智能配电网柔性负荷聚类分析方法。通过集成树拟合,实现智能配电网台区柔性负荷高维数据的降维。利用马氏距离克服聚类指标维度的相关性,进而确定有效聚类数。通过卡林斯基-哈拉巴斯指数(Calinski-Harabaz, CH)确定基准聚类算法。通过一致性函数矩阵对聚类结果进行统一,攻克了聚类结果不稳定的问题,得到更一致、更稳定的总体聚类结果。通过宁夏回族自治区智能配电台区实际运行数据分析,验证了所提基于投票策略的聚类算法的有效性。

In response to the significant load characteristic variations within intelligent distribution network stations and challenges in balancing clustering effectiveness,accuracy,and robustness of traditional clustering algorithms,this paper presents a clustering method for flexible loads in intelligent distribution networks based on a voting strategy.Combining the K-means algorithm with the fuzzy C-means clustering algorithm based on grey wolf optimization(GWO-FCM),the clustering stability and effectiveness are considered.To reduce the dimensionality of high-dimensional data related to flexible loads in intelligent distribution networks,an integrated tree-fitting approach is employed for dimensionality reduction of high-dimensional data pertaining to flexible loads.The Mahalanobis distance is used to overcome the correlation among clustering indices and determine the optimal cluster number.The benchmark clustering algorithm is determined based on the Calinski-Harabaz(CH)index.A consistency function matrix is used to unify the clustering results,mitigating instability in clustering outcomes and yielding a more consistent and stable overall clustering result.Finally,the validity of the proposed clustering algorithm based on the voting strategy is verified through actual operation data from intelligent distribution stations in the Ningxia Hui Autonomous Region.