Anomaly detection and clustering‐based identification method for consumer–transformer relationship and associated phase in low‐voltage distribution systems. Issue 6 (24th December 2022)

• Record Type: Journal Article
• Authors: Chu, Zhenyue; Cui, Xueyuan; Zhai, Xingli; Liu, Shengyuan; Qiu, Weiqiang; Waseem, Muhammad; Aziz, Tarique; Wang, Qin; Lin, Zhenzhi
• Is Part Of: Energy conversion and economics. Volume 3:Issue 6(2022)
• Page Start: 392
• Page End: 402
• Publication Date: 2022-12-24
• Subjects: clustering by fast search and find of density peaks -- consumer–transformer relationship -- fast dynamic time warping distance -- local outlier factor -- low‐voltage distribution systems -- phase identification; Energy conversion -- Periodicals; Energy conversion; Periodicals; 621.3124
• DOI: 10.1049/enc2.12073
• Languages: English
• ISSNs: 2634-1581

Abstract

Abstract: The identification accuracy of low‐voltage distribution consumer–transformer relationship and phase are crucial to three‐phase unbalanced regulation and error correction in consumer–transformer relationships. However, owing to the rapid increase in the number of consumers and the upgrade of the feed lines for low‐voltage distribution systems, the timely update of the consumer‐transformer relationship and phase information of consumers is challenging. This influences the accuracy of the basic information of the power grid. Thus, this study proposes a low‐voltage distribution network consumer–transformer relationship and phase identification method based on anomaly detection and the clustering algorithm. First, the improved fast dynamic time warping distance based on the filter search between voltage sequences is used to measure the similarity between voltage curves. Subsequently, an abnormal consumer detection method based on the local outlier factor is used to identify consumers with mismatched consumer‐transformer relationships by determining the local outlier factor scores of voltage curves. Furthermore, the phase information of normal consumers is identified through clustering by fast search and find of density peaks. Finally, the proposed method is validated using case studies of practical low‐voltage distribution systems in China. The proposed method can effectively improve phase identification accuracy and maintain high adaptability in various data environments.