基于强化学习思想的热风炉空燃比控制

AIR FUEL RATIO CONTROL OF HOT BLAST STOVE BASED ON REINFORCEMENT LEARNING

热风炉是高炉生产中的重要设备之一,也是耗能的主要设备。在节能降耗的大背景下,采用动态优化的空燃比组织燃烧,可以较好地达到节能减排的目标。强化学习是一类具有自学习功能的新型人工智能方法,越来越多地应用在自动控制领域。本文提出了一种基于强化学习思想的空燃比调节方法,通过时间循环神经网络建立热风炉燃烧过程模型,利用"动作决策模块"随机调整燃料阀与空气阀,根据合理的"评价体系"评估动作好坏,得出较为理想的空燃比。仿真结果表明,该方法可以模拟空燃比等变量与拱顶温度以及废气温度之间的关系,能够降低燃耗,减少人为干扰,提高经济效益。

Hot blast stove is one of the important equipment in blast furnace production,which is also the main equipment of energy consumption.Under the background of energy saving and consumption reduction,adopting the dynamic optimized air fuel ratio can save energy and reduce emission.Reinforcement learning is a new type of artificial intelligence method with self-learning function,which is more and more used in the field of automatic control.In this paper,an air-fuel ratio regulation method based on reinforcement learning is proposed.The combustion process model of hot blast stove is established through time cycle neural network,and the fuel valve and air valve are adjusted randomly by " action decision module".According to the reasonable " evaluation system",the better air-fuel ratio is obtained.The simulation results show that the method can simulate the relationship between the variables such as air-fuel ratio,vault temperature and exhaust gas temperature,reduce fuel consumption,reduce human interference and improve economic benefit.