Abstract: In view of problems of complicated disturbance of power quality of coal mine grid and selection of wavelet base when using wavelet analysis technology to solve transient process of disturbance, the paper proposed a detection method of starting and ending time of disturbance based on HHT. The method takes voltage sag as example, firstly uses EMD algorithm to decompose the voltage sag signal to obtain IMF1 component, IMF2 component, IMF3 component and surplus component R under noise environment of 30 dB and 40 dB; then makes Hilbert transform for signals including high-frequency part namely original signal, IMF1 component and IMF2 component and gains a conclusion that starting and ending time of disturbance can be detected by position of catastrophe point on instantaneous amplitude curve; finally selects instantaneous amplitude of IMF1 component with the minimum error to detect starting and ending time of disturbance through comparing detection values of starting and ending time of disturbance of original signal, IMF1 component and IMF2 component. The experiment result showed that the method has high precision and strong anti-noise ability.