In this paper, the photovoltaic (PV) power prediction model is optimized according to the characteristics of PV output units in distributed energy industrial parks to provide data support for the subsequent dispatching strategy. The EMD-SARIMA forecasting model is a combination of Empirical Mode Decomposition (EMD) and Seasonal Autoregressive Integrated Moving Average (SARIMA). In the model, the problem of determining the period of each IMF component of the signal component is proposed, the period T calculation method incorporating fast Fourier transform (FFT) is proposed, and the obtained period is fed into SARIMA as an input parameter together with the IMF sequence for prediction, which constitutes the EMD-FFT-SARIMA prediction model. Then, the prediction results corresponding to each IMF are superimposed and reconstructed to obtain the final prediction results. The error calculation of the prediction results reveals that the root mean square error (RMSE) decreases from 120.6 MW to 19.3 MW, and the mean absolute error (MAE) decreases from 52.87 MW to 12.3 MW.