Taking the no-load air gap flux density waveform distortion rate of a 48 V permanent magnet synchronous generator (PMSG) for a certain range extender as the optimization object, using the permanent magnet pole arc coefficient, air gap length, magnetic steel thickness, and magnetic pole eccentricity as variables, and the magnetic steel quality and the generator efficiency as the evaluation reference indices, the orthogonal experiment matrix is established by the RMxprt parameterized scanning function and the Taguchi method. The data samples are obtained by the ANSYS Maxwell two-dimensional finite element simulation analysis. Then, the sample data are fitted by the Fourier model to construct the fitness function. Finally, particle swarm optimization (PSO) is used for optimizaiton, and a set of optimal magnetic steel parameters is obtained and verified by finite element method. The results show that using the Taguchi-PSO method, the no-load air gap flux density waveform is greatly improved and the mass of magnetic steel is reduced.