To forecast drone demand for “last mile” distribution， we consider many factors， including drone performance， airspace environment and transportation tasks. The objective function is to maximize the delivery volume and minimize transportation cost with a multi-stage unmanned aerial vehicle（UAV） demand forecasting model. Considering the historical express data and influencing factors， a combination forecasting method is used to solve the express delivery business volume. The dynamic allocation algorithm can calculate the drone delivery volume and obtain the optimal demand for the drone prediction. The actual express delivery data and low-altitude flight conditions in a certain area are analyzed as examples. The results show that the proposed prediction method can provide flexible demand schemes according to the requirements of the distribution center for unmanned aerial vehicles， working time and transportation costs.