Coal injection rate adjustment is the main way and means to control the temperature of each section in the rotary kiln, but it is difficult to effectively control the temperature of each section due to such characteristics as nonlinearity, time delay, multivariate and uncertainty of the calcination process of pellets in the rotary kiln. A prediction model of coal injection rate in rotary kiln based on particle swarm optimization BP neural network (PSO-BP) is constructed, and the predicted value and actual value of coal injection rate are compared through simulation experiments, and the error and fitting effect of the identification model are analyzed. The results show that the precision of the PSO-BP neural network prediction model is better than that of the original BP neural network, with high optimization efficiency and good overall fitting effect, with a linear fitting degree of about 0.95, an absolute error of about 0.09 t/h, and a relative error of about 2.4%. Based on the PSO-BP neural network model, the average relative error of the temperature prediction section is 0.8%, and the overall average temperature fluctuation is within ±10 ℃. It shows that the constructed prediction model of coal injection rate in rotary kiln is accurate, reliable and applicable, and the established neural network is combined with the traditional PID control to form an automatic temperature control system for the preheating section, which can realize the stable control of the temperature field of the rotary kiln.