Load identification based on optimized fuzzy C-means state extraction

Abstract

With the increasing consumption of electric energy in China, the management of power consumption becomes more and more important. It is essential for the analysis of electric appliances to make a reasonable way of electricity consumption. Aiming at the problem of low accuracy of low frequency sampling recognition, this paper proposes a non negative matrix decomposition method for feature extraction of electrical data. The peak density of feature data is calculated by peak density algorithm, and the peak density of feature data is taken as the initial cluster center of fuzzy c-means algorithm. The feature States of each device are obtained. Finally, the state sequence is processed by single and two-way long-term and short-term memory network. The accuracy of load identification is greatly improved. It shows that the proposed method has outstanding capability for device state extraction.