基于DLT-Kmedoids算法的用电负荷聚类分析

Power Load Clustering Analysis Based on DLT-Kmedoids

针对高校用电负荷中传统聚类算法直接应用于时间序列聚类效果准确性较低的问题,提出一种融合DTW距离、LB_Keogh距离以及时间窗口的DLT-Kmedoids算法,以提高聚类算法应用于时间序列的准确性以及算法效率。DLT-Kmedoids算法使用DTW计算时序数据之间的距离取代传统的欧氏距离度量方式,提高了相似性度量算法精度,同时也提高了聚类算法的准确性和复杂度,引入LB_Keogh距离在计算DTW距离之前过滤掉大部分不可能是最优匹配序列的序列,对于剩下的序列再使用DTW逐个比较,进一步降低算法的复杂度。最后结合高校建筑用电负荷时间序列数据进行分析,通过与主流聚类算法进行比较,表明该算法对于高校用电负荷数据的聚类任务,能够更准确地识别相似的负荷模式,并以更高的效率进行聚类分析。

Aiming at the problem of low accuracy in directly applying traditional clustering algorithms to time series clustering in the electricity load of universities,a DLT-Kmedoids algorithm combining DTW distance,LB_Keogh distance and time window is proposed to improve the accuracy and efficiency of clustering algorithm applied to time series.The DLT-Kmedoids algorithm uses DTW to calculate the distance between time series data instead of traditional Euclidean distance measurement,improving the accuracy of similarity measurement algorithms and also improving the accuracy and complexity of clustering algorithms.LB_Keogh distance is introduced to filter out most sequences that are unlikely to be optimal matching sequences before calculating DTW distance,and DTW is used to compare the remaining sequences one by one to further reduce the complexity of the algorithm.Finally,we analyze the time series data of electricity consumption in university buildings,and compare it with mainstream clustering algorithms.It is showed that the proposed algorithm can more accurately identify similar load patterns and perform clustering analysis with higher efficiency for the clustering task of electricity consumption data in universities.