Feature extraction of electric information acquisition system based on Haar wavelet transform

This paper studies the structure features of the collected power consumers' power consumption data, in which the 96 measured data points indicating a single user's power consumption curve of a day. The 96 measured data points are taken as 96 variables for analysis. The dimensionality of original data is reduced by principal component analysis (PCA), and then the K-means algorithm is employed to cluster the original time-series load data of power consumers. In detail, Haar wavelet transform is adopted to provide a five-level multi-resolution decomposition of the original load data, from which 93 Haar coefficients and 3 scale coefficients are obtained. The cluster result is taken as category labels, and the standardized 93 coefficients and 3 scale coefficients as dependent variables, a classification model based on C5.0 algorithm is established to confirm the Haar scale coefficients as characteristic quantity of power consumption curve. Thus, the data acquisition is simplified and storage space compressed. This method could also be adopted for feature extraction of tremendous detail data and indicator data in power system, with potential application value and economic benefit.