A　variety　of　pulses　and　white　noises　exist　in　the　low-voltage　distributed　systems.　These　noises　are　obstructive　to　the　accurate　extraction　of　the　short-circuit　faults.　The　effective　filter　ought　to　be　designed　to　utilize　the　abrupt　change　of　the　current　deviation　in　the　fault　occurrence　for　the　purpose　of　the　early　fault　detection.　In　this　way,　the　characteristics　of　the　fault　signal　can　be　preserved　and　the　noise　interferences　are　suppressed.　The　morphological　filter　is　utilized　as　the　preposed　unit　of　the　multi-scale　wavelet　transform,　which　are　combined　as　a　novel　model　of　the　morphology-wavelet　filtering　algorithm　and　applied　to　the　ultra-high　speed　detection　of　the　short-circuit　faults　of　the　low-voltage　systems.　This　model　incorporates　the　advantages　of　morphological　filter　and　multi-scale　wavelet　transform　to　extract　the　feature　of　faults　meanwhile　restraining　various　noises.　Besides,　it　can　be　implemented　in　real　time　using　the　available　hardware.　The　effectiveness　of　this　model　was　verified　with　the　simulation　results　and　the　tests　based　on　TI　TMS320F2812　DSP.