一种面向需求响应资源的模糊聚类算法

A Fuzzy C-Means Clustering Algorithm for Demand-Side Response Resource

传统的依据工、商、居民等信息的资源分类方式,难以获得较高的资源响应达标率,降低了用户体验和系统效率。该文提出面向需求响应资源的聚类算法。该算法首先利用傅里叶级数,选取五维低频分量,进行负荷信息的压缩和提取。然后结合资源的出力特性、响应特性等,用向量形式表示需求响应资源,构建了响应资源的表示空间。进一步的,基于模糊C均值的资源聚类算法,对资源进行聚类。最后,算法对系统已执行的需求响应效果进行充分挖掘,利用老化函数、响应达标情况的评分,针对两种不同的达标情况表示方式,实现了资源聚类的修正。实验验证表明,该文所述算法,可以实现资源的聚类,从而显著提升需求响应的效率,提高资源的达标率。

It＇s difficult to achieve the goal of demand-side response（DR）, based on the traditional classification such as industry, business, residential load and so on. The low success rate of DR resource leads to poor user experience and low system efficiency. A new clustering algorithm for DR resource was introduced in this paper. The information of load was extracted in the way that lowest five frequency component was remained via Fourier series. Then, the vector of DR resource was built with the resource character of load and response ability. What＇s more, to deal with the problem related with data variance, the vector was unitized to form a space, in which the distance was also defined. Then the resource was classified by FCM algorithm. At last, the relationship between resource and DR was detected, and two method of amending classification were put forward considering aging function, success rate of DR, and the difference between two representation ways of DR success. Jiangsu＇s pilot program shows that the clustering algorithm in this paper performs well, the DR success rate of resource is improved significantly.