Typical Load Patterns Clustering for Low-Voltage Customers Based on Parallel-Teda Algorithm

Abstract

The widespread deployment of smart meters has facilitated the development of big data analytics in the power distribution sector. However, the data generated from low-voltage clients' electricity consumption is continuously increasing and becoming more complex over time. To address the challenges posed by the large-scale and high-dimensional nature of this data, this paper proposes an innovative parallel data analysis framework, named Parallel-Typicality and Eccentricity-based Data Analysis (TEDA), which combines parallel computing with the TEDA theory. This entirely data-driven framework is designed to be adaptable to evolving data patterns and is free of prior assumptions. The Parallel-TEDAframework is capable of processing data at significantly faster speeds than conventional parallel processing methods. Moreover, it can efficiently avoid meaningless fusion problems by utilizing the advantages of TEDA. The proposed algorithm is tested on a real-world dataset of low-voltage customer loads. Simulation results show its superiority in terms of clustering quality and computational efficiency.