This paper proposes a high impedance fault detection method based on particle swarm optimization algorithm and Bayes classifier. This method uses the discrete wavelet transform to build the time-frequency matrix about distribution network voltage and current, to extract the characteristic attribute of high impedance fault. Particle swarm optimization (PSO) is used to optimize the feature space of Bayes classifier to improve the classification accuracy and computational timeliness. The simulation and experimental results show that the classification accuracy of proposed method is more than 95%, it can effectively deal with interference factors such as insulator leakage current, the capacitor switching and nonlinear load.