Power Quality Time Series Clustering Analysis Based on Symbolic Feature Representation Method

In order to solve the problem of low accuracy and high complexity in traditional power quality data processing methods, a symbolic feature representation method is proposed to analyze the similarity among power quality time series (PQTS). The characteristics of PQTS are extracted through recursive principal component analysis method, which is used to construct fast recursive expressions. The symbolic feature representation is established to reduce data dimensions by using piecewise linear fitting method. The distance among the symbolic feature representations is measured to accomplish the similarity clustering of time series. The analysis result of simulation case indicates the advantage of accuracy and effectiveness of the proposed method.