At present, the power grid companies of China basically adopt expired replacement and batch sampling rotation for smart meters, which causes a waste of human and material resources. This paper proposes a method for identifying and recommending that the operating error of a smart meter is out of tolerance based on the electricity consumption data of the metering automation system. Firstly, the topology and mathematical model for analyzing the operating error of the smart meters under the distribution transformer are present. Secondly, considering the ill-conditioned problem, a solution method based on Tikhonov regularization （Tik) and generalized cross-validation （GCV) to optimize the regularization parameters is suggested. Then, a sorting recommendation mechanism for identifying out-of-tolerance smart meters is proposed. Finally, it is verified with actual distribution grid data in a certain district and compared with other methods. The results show that the proposed method is practical and effective.