Electricity consumption patterns analysis using a shape-based clustering method based on seasonal decomposition and parallel computing
Kaiwen Li, Tao Zhang, Jun Meng, Dan Li, Hongwei Che, Dudu Tan · 2019
There is a significant difference in the electricity consumption patterns of different users. Recognizing and extracting the electricity consumption patterns can help analyze the user behaviors and support the market decision-makings. Clustering algorithms have proven to be an effective technique to identify the load patterns. In this work, we propose a shape-based clustering method based on seasonal decomposition and parallel computing to identify the electricity consumption patterns in Changsha residential areas. The seasonal decomposition is first used to extract the representative curve which serves as the inputs of the clustering. A shape-based clustering algorithm, which better considers the shape characteristics of load curves and requires less computational cost, is introduced to partition the load data into different groups with different electricity consumption patterns. In addition, the parallel computing is employed to further accelerate the computing time. The experimental results demonstrate that our method can identify different electricity consumption patterns effectively. Moreover, our method surpasses the traditional DTW-based and fuzzy clustering algorithm and is 30 times faster than the DTW-based clustering algorithm. At last, we analyze the electricity consumption patterns in the Chinese Spring Festival in Changsha residential area that further validate the efficiency of our method.