An Adaptive Gradient Descent Method for Error Estimation of Electric Meters
Error verification of electric meters in the power industry is usually manually conducted through standard meters. With the continuous improvement of real-time data collection technology, data of power systems available for analysis is becoming more abundant. In this paper, we propose an adaptive gradient descent method for error estimation of electric meters based on large amount of data. In order to improve the accuracy of estimation results, we first adopt a clustering algorithm for light load data detection and elimination. Then we provide a detailed description of the remote estimation model for the running error of electric meters. According to the simulation experiments, results obtained by the proposed method can well match the true value of electric meter's running error. This method can effectively reduce the maintenance cost of on-site calibration of electric meters, and can also provide a reference for the service of electric meters.