Parameters inversion of partial differential equation from observed data has great values. Based on FFT spectral method and Nelder-Mead simplex algorithm,a practical algorithm is presented. Given the guess values of parameters,the prediction values of partial differential equations solutions can be computed by the FFT spectral algorithm,and the optimization objective function can be constructed u-sing difference between observed data and prediction values.So,the optimal parameters and order are searched by Nelder-Mead simplex optimization method.Finally,by computer simulation,a set of feasible inversion method is presented for study scientific and engineering problems.