Abstract:

The traditional load curves clustering method is generally based on Euclidean metric,but the curves with different shapes can have the equal Euclidean metric relative to the same clustering center.Therefore,the Euclidean metric based clustering method can not accurately classify the curves in morphology.Aiming at the this shortcomings,an adaptive shape clustering algorithm based on discrete curves variance comparison is proposed.The proposed method takes the variance of the difference between the load curve and the clustering center as the basis,to realize the classification based on the curve shape.In this paper,the principle and shortage of Euclidean metric based clustering method are analyzed,then the basic principle of shape clustering based on variance comparison and its corresponding clustering center updating method are detailed related.Based on the simulation and the actual examples,theses two clustering method are analyzed and compared with each others.The results show that the variance comparison based clustering algorithm has better performance on shape classification,the feasibility and validity of the proposed method are verified by the final analysis results.

Key words: electric load, cluster analysis, variance, curve shape, Euclidean metric