Abstract: One kind of detection and prediction method for abnormal grid voltage has been designed due to the case that modern power electronic equipment is sensitive to non-stationary time-varying voltage signal. This method sends the network voltage abnormal warning signal to control-circuits of modern equipment through detecting the grid voltage in time. To eliminate the conventional noise jamming of the voltage signal,this scheme adopts linear tracking differentiator to filter the signal. On this basis,wavelet transform modulus maxima are proposed in singularity detection,so as to accurately forecast abnormal harm points in power electronic devices caused by the grid voltage. Simulation and experimental results show that the wavelet analysis based on linear tracking differentiator can obtain the best approximation of the ideal signal and provide more useful forecasting signals for power electronics equipment,thus the fault detection speed and efficiency are improved significantly.