Abstract: In order to scientifically identify and judge the trend of local COVID-19 recurrence， and to control and prevent the spread of its epidemic， a novel time-divided SIQR coronavirus pneumonia infection model with nonlinear parameters is established. It divides the epidemic spread into three periods， and comprehensively considers the intervention effect of public opinion and prevention measures on each period. The simulation results on several real datasets show that the rapid enhancement of the screening of infected and uncontrolled people will significantly interfere with the spread of the epidemic and make the inflection point of slowing down the number of infections come early. The prediction results of the epidemic situation at Dalian in December 2020 by the proposed model not only have smaller mean square error， but also the change trend of the predication curve is closer to the reality compared with other baseline models. Additionally， the epidemic situation with longer free transmission period and stronger aggregation is simulated by the proposed model. The prediction error is also smaller than that of the compared models on the real data. This shows that the model has certain universality and robustness.

Key words: COVID-19; epidemic model; public opinion factors; simulation; prediction