The new crown epidemic spreads and spreads rapidly in various regions along with the movement of people and mutual contact. An epidemic spread model (D-SEIR model for short) that considers the population flow among regions and within regions was proposed. Through comprehensive consideration of the flow of people, and the changes in the number of susceptible and latent people, the changes in the confirmed population can be predicted. The experimental results of empirical data in Shenzhen and Shandong show that the proposed D-SEIR model can accurately fit the evolutionary trends of susceptible and diagnosed populations during the transmission of COVID-19. Prior to the change in the epidemic situation, the goodness-of-fit ${{{r}}^2}$ of the change trend of the susceptible and the confirmed population in Shenzhen reaches 0.9661 and 0.8474, The fitness ${{{r}}^2}$ of Shandong province reaches 0.8576 and 0.8795. With the stabilization of the epidemic situation, the goodness-of-fit $ {r}^{2} $ of Shenzhen reaches 0.9720 and 0.8180. The fitness ${{{r}}^2}$ of Shandong province reaches 0.8615 and 0.9122. The model of this paper is helpful to understand the propagation rules and characteristics of COVID-19 under the flow of people, and provides decision-making basis for early risk warning of different cities and regions under the background of resumption of working and production.