Abstract: In allusion to the problems of low gas recovery rate and big smoke of traditional gas recovery system,the paper put forward a predictive control strategy of gas recovery system based on neural network,which optimizes gas recovery system of a steel plant’s converter by applying neural network adaptive and predictive control and fuzzy control.The simulation results showed that the predictive error of furnace gas emissions was-5～5 L/h and the preditive effect was better.The practice application proved the average recovery of gas reached 97.5 m3/t after applying neural network adaptive and predictive control and it reached the purpose of energy saving,low cost,and protection of environment.