Aiming at the problem that the current comprehensive error calculation method of electric energy metering can not evaluate the error of electric energy metering device in real time, an error evaluation algorithm of electric energy metering device based on PCA-SVR is proposed. The method first carries out principal component analysis (PCA) on the effective value of secondary signal of transformer, and detects the measurement state and locates the anomaly of the electric energy metering device by Q statistics and its contribution rate. Then, a multi-parameter dimension reduction model of synthetic error of energy metering device under normal metering state is established. The measurement error under actual working condition of transformer is obtained by support vector machine regression (SVR). The comprehensive error is synthesized with secondary circuit error and watt-hour meter error obtained by on-line monitoring. This method can realize the real-time state evaluation and synthesis error evaluation of energy metering errors. Finally, the accuracy of this method is verified by simulation.