Aiming at the problems of high labor cost, poor efficiency, and small coverage in the current error verification of smart meters, an out- of- tolerance smart meter screening method based on line loss iteration is proposed. Firstly, the daily electricitydata of the distribution station is collected and screened over a period of time to identify the data of light and no-load meters. Then,the fixed loss, line loss and error coefficient of meters are calculated iteratively by using the error estimation model. Through iteration, the data with high and low line loss rate are gradually eliminated, so that the line loss rate of effective data is concentrated inan interval. This not only improves the accuracy of error estimation, but also prevents valid data from being deleted by mistake. Finally, a non- light and no- load meter whose error coefficient exceeds the standard value is labeled as an out- of- difference one.The validity of the proposed method is verified by the data of electricity meters in three different sizes of distribution stations, and the experimental results show that the proposed method can detectout-of-tolerance electricity meters accurately. Comparing with the traditional least-squares method, the detection accuracy of the proposed method is obviously better than it.