Pattern Recognition of Traction Energy Consumption for Urban Rail Transit by Using Symbolic Aggregate Approximation

Abstract

In the urban rail transit system, the traction energy consumption accounts for 40%-60% of the total energy consumption. There is a large amount of traction energy consumption data in time series format recorded by energy meters. Accurate analysis of traction energy consumption based on time series is in urgent demand for energy saving. In order to analyze the law of traction energy consumption, this paper proposes a pattern recognition method for traction energy consumption based on SAX (Symbolic Aggregate approXimation). The original time series of traction energy consumption is transformed by SAX and the sub-patterns are obtained. The traction energy consumption patterns are recognized by using K-means algorithm. To show the effectiveness and efficiency, we apply the proposed method to a data set from Beijing Subway, and find 3 representative patterns. We find that the recognized patterns of traction energy consumption appears coherence with the major services prescribed in the rolling stock scheduling plan. By calculating the similarity and comparing with these representative patterns, the days that differ from the typical patterns are judged as anomalies.