At present, the main problem of electronic voltage transformer is the degradation of accuracy after long-term operation. The current methods include regular off-line calibration and online monitoring. The former cannot find the change of the transformer error in time, while the latter needs the standard to be grid-connected for a long time, which cannot be applied on a large scale. Thus, an error prediction method of electronic voltage transformer based on transfer entropy and wavelet neural network is proposed in this paper. Firstly, the main influ-encing factors of the ratio difference and the angle difference are selected by transfer entropy. Then, the factors are used as the input, and the error prediction model of the wavelet neural network is established and tested. The simulation show that the prediction error of the ratio difference is less than 5%, and the prediction error of the angle difference is less than 10%. The proposed method can realize the condition monitoring of the transformer for a long time.