【Purposes】 Manual inspection’s lengthy time commitment, poor efficiency, and inaccurate verification make it challenging to meet practical application requirements in the daily operation of electric energy metering devices. 【Methods】 According to the collected data of electricity information, an evaluation model of metering device operation status based on deep learning is established, and the characteristics of historical data of electricity through the deep learning model are captured. By using the transfer learning optimization model training process, the prediction of the users’ future electricity usage is completed, and the threshold value is set for the difference between the expected value and the measured value of electricity to judge the running state of the energy meter. 【Findings】 The study’s findings will be a solid foundation for assessing the regional electric energy metering system’s state of operation, as well as for accurately replacing and maintaining meter. In addition, it will increase the precision of power grid fault diagnosis and location while significantly lowering the cost of on-site maintenance. 【Conclusions】 By contrasting the simulated value with the actual platform value, the correctness of the method is demonstrated. The experimental results show that the proposed evaluation model is effective.