Residential Customer Clustering Based On Household Electricity Load Disaggregation

Abstract

Customer clustering is important for understanding electricity usage behaviors of residential customers and designing residential pricing mechanism. This paper aims to develop an efficient customer clustering method by uncovering the temperature-demand dependence of each customer. Specifically, we use a load disaggregation approach to extract several representative parameters related to heating and cooling season characteristics. These parameters can be used as input features to group customers through K-means clustering. Real-world load data tests have identified clusters of several unique features, such as house sizes, the ownership of large appliances, or similar temperature-based electricity use behaviors.