A missing data complement method based on K-means clustering analysis

This paper proposes a missing data complement method based on k-means clustering analysis by stretching and replacing energy load curve in proportion. The historical load data is classified in units of daily load vector considering the cycle similarity of energy load. The centroid vector of the corresponding clustering curve is stretched in proportion, and replacing the missing data. Simulation results show that the proposed missing load data complement method is more accurate compared to traditional linear interpolation method, and is able to improve the load prediction accuracy to a certain extent worked with support vector machine (SVM).