基于灰狼优化聚类算法的日负荷曲线聚类分析  

Clustering analysis of daily load curves based on GWO algorithm

作　　者：吴亚雄 高崇 曹华珍 陈吕鹏 唐俊熙 李浩 WU Yaxiong;GAO Chong;CAO Huazhen;CHEN Lüpeng;TANG Junxi;LI Hao(Grid Planning&Research Center,Guangdong Power Grid Co.,Ltd.,CSG,Guangzhou 510030,China;Guangdong Power Grid Development Research Institute Co.,Ltd.,Guangzhou 510030,China;Suzhou Huatian Power Technology Co.,Ltd.,Suzhou 215000,China)

机构地区：[1]广东电网有限责任公司电网规划研究中心,广东广州510030 [2]广东电网发展研究院有限责任公司,广东广州510030 [3]苏州华天国科电力科技有限公司,江苏苏州215000

出　　处：《电力系统保护与控制》2020年第6期68-76,共9页Power System Protection and Control

基　　金：国家自然科学基金项目资助(51777078);中国南方电网公司科技项目资助(GDKJXM20172939)。

摘　　要：针对模糊C-均值聚类算法(Fuzzy C-Means,FCM)应用于日负荷曲线聚类分析时存在易受初始聚类中心影响,易收敛于局部最优值以及日负荷曲线的内在特性难以通过距离得到充分反映的问题,利用日负荷特征值指标对日负荷曲线进行数据降维处理。提出了基于灰狼算法(Grey Wolf Optimizer,GWO)优化的模糊C-均值聚类算法(GWO-FCM)。该算法利用GWO为FCM优化初始聚类中心,结合了GWO的全局搜索能力和FCM的局部搜索能力。算例结果表明所提方法可有效提高日负荷曲线聚类效果,算法鲁棒性好。Fuzzy C-Means(FCM)is susceptible to the influence of initial clustering centers,easy to converge to local optimum values and the inherent characteristics of daily load curve can not be fully reflected through distance when applied to clustering analysis of daily load curve.In order to solve the above problems,the dimensionality of daily load curve is reduced by using the characteristic value index of daily load.A fuzzy C-means clustering algorithm based on Grey Wolf Optimizer(GWO)optimization is proposed.The algorithm uses GWO to optimize the initial clustering center for FCM,and combines the global search ability of GWO and the local search ability of FCM.The results show that the proposed method can effectively improve the clustering effect of daily load curve and has good robustness.

关 键 词：日负荷曲线聚类分析 灰狼优化算法 模糊C-均值聚类算法 数据降维 算法鲁棒性 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论] TP18[自动化与计算机技术—计算机科学与技术]