The invention discloses a fractional order state spatial model identification method for a proton exchange membrane fuel cell. Firstly, hydrogen flow and load current are selected to serve as input variables, voltage is selected to serve as an output variable, a large amount of data is collected, and a Poisson moment function is used for wave filtering of the data; secondly, based on the Grunwald-Letniko fractional order calculus definition, a short-term memory matrix and an input-output matrix are constructed, and a matrix projection is constructed according to the result of a matrix function; thirdly, the SVD singular value decomposition of the projection is calculated, the system order is determined, an observable matrix is obtained, and according to the observable matrix, solution of system matrixes A, B, C, and D is conducted; finally, solution of a cost function is conducted, and the optimal system fractional order is obtained. The prior knowledge of an electric pile model is notrequired, the identification speed is high, the identification accuracy is high, the universality is good, and the method facilitates subsequent controller design and simulation.