The Global Prediction System of the COVID-19 Pandemic (GPCP) has reliable predictions over the number of daily and seasonal new cases of COVID-19 in each country based on updated epidemiological data in real time. The first version of the model uses a modified epidemiological SIR model combining authentic data of global pandemic and considering impacts of the meteorological factors and quarantine measures on the spread of COVID-19. The second version of our model uses a more complex modified SEIR model. In this version, we take into account impacts of the community reopening time and citizen self-quarantine on the development of epidemic. The second version of the model can be used to make seasonal predictions and predictions of the second outbreaks. The parameters of the model are obtained by inversion of authentic epidemiological data. At the same time, we adopt the method of EEMD-ARMA to modify the prediction results to obtain a better prediction effect.