Load peak-valley shape clustering and drift analysis for improving temporal pattern representation

Load shape pattern clustering is an important foundation for developing appropriate tariff design and load management to achieve more economical and reliable benefits. However, the existing load shape pattern clustering methods mainly focus on the whole load shape and various clustering algorithms, which do not consider the peak-valley shape features and distribution drift issue of the load shapes. Therefore, peak-valley shape pattern clustering and drift measurement of daily load shapes are proposed to solve this problem. To accurately reveal the peak-valley electricity consumption behaviors, load peak-valley shape models and a hybrid distance measurement are proposed to obtain more representative temporal patterns that have more compact peak-valley shape distributions. Moreover, two measurement models for power drift and time drift are proposed to analyze the significant drift problem between daily load peak-valley shape patterns. The results showed that the proposed method outperformed other methods, as it not only achieved the best clustering effectiveness scores, such as DBI, WAS, CHI, SC, and DI scores of 0.5537, 0.2633, 502.3634, 0.8872, and 1.4730, respectively, but also accurately obtained the time drift values between different modes, such as the maximum backward shift of the two peak times by 90 and 150 minutes, and the maximum backward shift of the valley time by 45 and 60 minutes, respectively.