A Novel Load Clustering Method Based on Entropy Features Considering Longitudinal Characteristics

Customer behaviors in power consumption can be described by their transverse and longitudinal characteristics. Transverse characteristics can be expressed using daily load curves, while longitudinal characteristics reflect the difference between daily load curves everyday. The traditional load clustering methods only consider the transverse characteristics of customer behaviors in power consumption, and don't take the longitudinal characteristics into account. To tackle this problem, a new load clustering method considering both the transverse and longitudinal characteristics of customers' power consumption behaviors is proposed. The entropy is used to characterize the longitudinal volatility of customers' power consumption behaviors. A group of household power load as a benchmark example is employed to analyze the proposed method, which shows that the proposed method is more appropriate than the traditional clustering methods which only consider the transverse characteristics. Finally, the entropy weights are utilized to quantify the demand response potential of each type of customers.