Abstract: In order to effectively utilize the fault characteristic of current surge after the occurrence of a short-circuit fault to achieve accurate identification of short-circuit faults,an effective filter must be designed to retain the fault feature while weakening the influence of noise.In this paper,a combined average generalized morphological filter was proposed to study its filtering effect on impulse noise,high frequency noise and Gauss white noise,as well as the improvement of the accuracy of the wavelet decomposition method.It is concluded that the combined average generalized morphological filter can effectively suppress the above three kinds of noise and significantly improve the accuracy of the detection algorithm.By comparing the combined average generalized morphological filter with the median filter and mean filter,the conclusion is drawn that the combined average generalized morphological filter has better lifting effect.The experiments show that the combination of combined average generalized morphological filter and wavelet decomposition method can make the lifting coefficient reach 11.21,the detection time costs 0.35 ms,and the filter delay only takes 0.11 ms.This combination has the real-time and rapid characteristics of fast detection of short-circuit faults,and is suitable for fast detection and fault protection applications of short-circuit faults in lowvoltage distribution systems.

Key words: generalized mathematical morphology, combination filter, denoising, short circuit detection, low voltage system