Abstract：The number of new energy vehicle charging facilities is increasing, and the accuracy and reliability of power measurement are directly related to the users' own interests. However, due to aging, failure and other factors, the performance of smart meters usually decreases over time while on-site verification requires a lot of manpower and material resources. Therefore, this paper proposes an error estimation method based on a high-speed convolutional neural network and bidirectional long short-term memory network. Firstly, the data characteristics of charging facilities collected by smart meters are preprocessed. Secondly, the spatial features between variables are extracted based on the convolution module added to the high-speed network, and part of the original information is retained. Finally, the proposed model is verified in the data set of an energy vehicle charging station, and compared with three existing traditional models, PSO-BPNN, EKF-LMRLS, GDRLS, as well as two single neural network models, CNN and LSTM. The experimental results show that the proposed method has significant advantages in error estimation accuracy of charging meters, and the three performance evaluation indicators have at least 13.68% improvement.
2025, Vol. 46 
