The fast and accurate detection of the magnitude

phase angle jump and the time instant of the beginning and end of voltage sags is the premise of sag compensation. We have proposed a new method of detecting voltage sags based on strong tracking filter（STF）. Firstly

a ten order STF model composed of DC

fundamental and harmonic components is founded in which the effect of harmonics

the frequency deviation

DC offset

non-integral harmonics and measurement noise has been taken into account. The fading factor of low order STF model is used to detect the time instant of the beginning and end of voltage sags. The experimental results show that this method is suitable not only for detecting deep sags but also for shallow sags. A high order STF model is used to detect the the magnitude and phase angle jump of voltage sags. The simulation using the synthetic signals shows this method has faster dynamic response characteristic than Kalman filter method and virtual three phases dq transformation method. The detection time is less than 0.2 ms

the amplitude error less than 0.009% and phase angle error less than 0.7 degree. It is found that the non-integral harmonics have little influence on this algorithm through selecting the smaller processing noise covariance Q and the larger measurement noise covariance R. Furthermore

this method has the superiority over the time-varying Kalman filter method since it is immune to frequency deviation and DC offset. The simulation using the real voltage sag event shows the proposed method can be applied in practice. It can be used to improve the DVR real-time compensating effectiveness.