基于多维缩放和KICIC的电力负荷聚类

电力负荷曲线聚类在电力大数据研究中有重要的应用。针对传统负荷聚类方法难以有效处理海量化的高维负荷数据，以及存在簇间样本模糊导致算法聚类质量不高、聚类效率低下等问题，提出一种结合多维缩放(multi-dimensional scaling, MDS)和一种新的集成簇间、簇内欧式距离的加权K-means方法(weighting k-means clustering approach by integrating intra-cluster and inter-cluster distances, KICIC)的聚类算法(MDS-KICIC)。该方法首先采用MDS算法对高维负荷数据进行数据降维处理，得到降维后的低维矩阵和归一化的特征值向量作为KICIC算法的输入矩阵和权重向量，KICIC通过在子空间内最大化簇中心与其他簇数据对象的距离来融合簇内和簇间的距离进行聚类，得到最终聚类结果。通过算例表明该方法运算时间短、聚类质量高，进一步提高了负荷曲线的聚类性能。

Power load curve clustering research based on Multi-Dimensional Scaling and KICIC

Power load curve clustering has important applications in the research of power big data. To address the problems that traditional load clustering methods are difficult to effectively deal with the increasingly massive and high-dimensional load data, and there are problems such as poor quality of algorithm clustering due to the ambiguity of inter-cluster samples and low efficiency of clustering, we propose a new weighting k-means approach combining multi-dimensional scaling (MDS) and a weighting k-means clustering approach by integrating intra-cluster and inter-cluster distances (KICIC) clustering algorithm (MDS-KICIC). The method first uses the MDS algorithm to reduce the dimensionality of the high-dimensional load data and obtains the reduced distance matrix and normalized eigenvalue vector as the input matrix and weight vector of the KICIC algorithm, and the KICIC clusters the intra-cluster and inter-cluster distances by maximizing the distance between the cluster center and other cluster data objects in the subspace to obtain the final clustering results. The arithmetic example shows that the method has short operation time and high clustering quality, which further improves the clustering performance of the load curve.