Entirely accurate electric metering data are needed in electricity market operations. According to the analysis of operating environment of tele-meter reading（TMR） system and traditional identification methods， a new identification method based on pattern recognition is proposed. Firstly， the acquisition process of TMR system is analyzed in detail. Then it is divided into several characteristic patterns， and to each pattern different index is used to identify the metering data. Finally， according to the analysis of all these indexes thoroughly， the data is proven to be true or false eventually. The data will be replaced by SCADA accumulated data once it is proven to be false. The method can improve the effectiveness of identification of electric metering data.