Analyzing seasonal variation in residential load patterns via two-stage clustering and relative entropy: poster

Residential load data generated by smart meters usually contain valuable information on household electricity consumption behaviors. Identifying load patterns and establishing typical consumer profiles will benefit stakeholders, including consumers, retailers, and operators. Load clustering has been widely studied in consumption segmentation and load profiling. However, it appears that the same consumers may vary their daily load patterns when the season changes, calling for comprehensive analysis into seasonal load patterns. This paper aims to study seasonal variations in load patterns. We model the seasonal load patterns using two-stage K-Medoids clustering and evaluate the relative entropy of the cluster distributions between seasons. Based on real-world load data, we find that 32% consumers hold similar load patterns in different seasons, whereas 14% consumers tend to vary their consumption habits when certain seasons come.