<p>As an indispensable part of the power system, the accuracy of low-voltage smart meter readings directly affects the fairness of power consumption measurement, energy management, and user electricity bill calculation. However, owing to the aging of the internal components of the meter, changes in environmental factors, and fluctuations in the power load, nonlinear errors often occur in the readings of smart meters. To correct these errors effectively, this study proposes a nonlinear time series correction method based on Transformer model. The combination function integrated into the Transformer model encodes its position, designs the structure of the Transformer model, analyzes the prediction process of modal distribution, sets a time window to limit modal characteristics, and fuses the reading error data results of low-voltage smart meters. Thus, the nonlinear time series characteristics of reading errors of low-voltage smart meters are extracted. The reading data of low-voltage smart meter are converted, and the obtained error data are processed by Transformer model to complete the measurement error correction. The experimental results show that the method in this study meets the practical needs of high-precision measurement and can identify obvious anomalies in the current measurement of low-voltage smart meters with the lowest error.</p>