The accuracy of watt hour meter depends on the distribution of large error data. This paper estimates the operation error of sub meter according to the measurement data of assessment table and sub meter in the station area, and introduces the pot model to measure the value at risk (VaR) of the over threshold data of sub meter operation error. The results show that this method can predict the probability of the operation error exceeding the limit under different thresholds and confidence levels, and then evaluate the accuracy of the meters in the station area, which provides a more reliable basis for preventing the risk caused by the meter out of tolerance.