Load Curve Clustering Method Based on HP Filter and Rescaled range analysis

Annual load series clustering is the basis of modeling and scenario generation of annual load series. In view of the shortcomings of existing clustering methods in terms of feature extractions on load fluctuation, a method of daily load series clustering based on HP filter and rescaled range analysis is proposed. Firstly, daily load series can be seen as a trend component plus a fluctuation component, the HP filter is used on load series spanning a season to obtain a random process with independent trend component and fluctuation component. Secondly, the concept of daily load related features is used to extract the features of the trend component. The features of the fluctuation component are extracted based on rescaled range analysis. Then, load series are clustered with fuzzy C-means with the total feature combined with the trend component features and the fluctuation component features. The annual load series of a country are clustered using the proposed method. Results shows that the method has certain advantages and can reflect the characteristics of daily load fluctuation to a certain extent.