The setting value of the coal injection rate in the blast furnace ironmaking process is usually derived from the experience of the furnace manager, which has certain blindness and ambiguity. In order to increase the rationality of the coal injection rate setting, this paper proposes an improved particle swarm Optimize the prediction model of blast furnace coal injection rate based on extreme learning machine. Based on the actual operating data of a blast furnace, this paper uses chaotic inertia weights and adaptive learning factors to improve the convergence of the particle swarm optimization (PSO) algorithm, and introduces the cross-mutation operation of genetic algorithms to improve the global optimality of the algorithm, and then uses the improvement The particle swarm algorithm optimized extreme learning machine (ELM) builds an extreme learning machine prediction model (IPSO-ELM) based on the improved particle swarm algorithm. Finally, based on the operating data of a certain blast furnace, the IPSO-ELM prediction model proposed in this paper is compared with the traditional particle swarm optimization extreme learning machine prediction model (PSO-ELM) and the extreme learning machine prediction model. The results show that the IPSO-ELM prediction model has higher prediction accuracy, can achieve accurate prediction of the change of the blast furnace coal injection rate, and has high industrial application value.