The key function of adaptive reclosing is to correctly identify fault nature and quickly capture the transient fault arc extinction time. Based on the analysis of the waveform complexity of fault terminal voltage after circuit breaker tripping under transient fault and permanent fault, this paper presented an adaptive reclosing overall implementation by combining local mean decomposition (LMD), approximate entropy and support vector machine (SVM). After getting the PF components of fault signal by using LMD decomposition, the approximate entropy of the first three PF components is calculated, which constitutes a three-dimensional feature vector as the input of SVM to identify fault nature and to capture arc extinguishing moment. Simulation results verify that this method can intelligently distinguish the transient fault from permanent fault, and capture transient fault extinction time with a strong anti-noise ability.