Abstract: Aiming at the problems of insufficient accuracy, inefficiency, and limited coverage of traditional smart meter for over-performing meter sampling methods, this paper proposes an over-performing smart meter identification method based on the LASSO regularization constrained sparse optimization model with convex relaxation. Firstly, based on the physical topology of the station area, through the validation of the traditional smart meter error estimation model, this paper finds that there is a linear relationship between the line loss power and the total power consumption, then, the first inspection error is introduced as a regularization penalty term, and the over-poor smart meter identification model based on the convex relaxation LASSO regularization with constraint sparse optimization model is proposed. Finally, the coordinate descent algorithm is used to determine whether there is an overdifferential meter in the station area and determine its overdifferential value. Compared with the restricted memory least squares method and the FMRLS algorithm combined with dynamic line loss, the method in this paper is closer to the real value in the estimation of the line loss rate, accurate in judging the location of the overdiffuse meter, and has higher accuracy in identifying the overdiffuse value.