In this paper, an on-line fault diagnosis method for proton exchange membrane fuel cells is proposed based on a fast electrochemical impedance spectroscopy measurement. Specifically, a fast electrochemical impedance spectroscopy measurement system is developed in order to reduce the measurement cost and time. The current pulse injection is used as the disturbance injection method, and the electrochemical impedance spectroscopy is obtained by the continuous wavelet transform and the maximum likelihood estimation. Then, the parameters of an improved equivalent circuit model are identified according to the electrochemical impedance spectroscopy, and three model parameters are selected as the features for fault diagnosis. On this basis, a multi-fault diagnosis algorithm is proposed based on the binary tree support vector machine classifier. Experimental results show that the proposed fault diagnosis method can realize accurate and fast on-line fault diagnosis for proton exchange membrane fuel cells.