A parametric bootstrap algorithm for cluster number determination of load pattern categorization. (1st August 2019)

• Record Type: Journal Article
• Authors: Luo, Xing; Zhu, Xu; Lim, Eng Gee
• Is Part Of: Energy. Volume 180(2019)
• Page Start: 50
• Page End: 60
• Publication Date: 2019-08-01
• Subjects: Load pattern categorization -- Parametric bootstrap algorithm -- Cluster number determination -- Cascade clustering; Power resources -- Periodicals; Power (Mechanics) -- Periodicals; Energy consumption -- Periodicals; 333.7905
• DOI: 10.1016/j.energy.2019.04.089
• Languages: English
• ISSNs: 0360-5442

Abstract

Abstract: The latest development of smart grid technologies gives rise to big load data and requires load pattern categorization (LPC). How to determine a precise cluster number and choose an appropriate clustering algorithm are critical and still remain challenging in LPC. In this work, we propose a novel parametric bootstrap (PB) algorithm to address the cluster number determination problem in load pattern analysis. The proposed PB algorithm is more robust against dimensionality of data and more applicable for the load demand data which is usually of high dimensionality. The PB algorithm is also general and independent of data type, resulting in a more precise cluster number determined than existing methods with little fluctuation. Moreover, an effective cascade clustering scheme is proposed to categorize load demand data and analyze load patterns, based on the PB algorithm and the K-means++ clustering algorithm. The results indicate the feasibility and the superiority of the proposed approach. Graphical abstract: Image 1 Highlights: A parametric bootstrap (PB) algorithm was proposed to determine cluster number. The method is particularly effective for the data in high dimensional space. A cascade clustering scheme was designed for load pattern categorization. PB algorithm incorporated with K-means++ was applied to real-life load data.