摘要
某钢铁企业炼焦生产过程中集气管压力为人工经验设定,难以根据不同工况进行实时调整,使得蝶阀容易工作在极值范围,无法达到期望生产目标,提出以降低焦炉能耗为优化目标,蝶阀开度范围和集气管压力工艺设定值范围为约束条件的稳态优化思想。首先采用基于线性回归与最近邻聚类学习算法的RBF神经网络方法建立焦炉集气管压力设定值动态模型,得到集气管压力设定估值,然后以降低能耗为目标函数,蝶阀开度范围和集气管压力工艺设定值范围为约束条件,采用改进的粒子群算法对目标函数寻优得到集气管压力的稳态优化设定值。实际应用结果表明,采用该方法获得的优化结果满足了企业的要求,取得了较好的工业应用效果。
In the coking plant production process of an iron and steel enterprise,the set-point of gas collector pressure is set according to human experience;it is difficult to adjust the set-point in real-time according to different conditions,so the butterfly valve cannot work properly and the anticipant production target can't be reached.In order to solve these problems,a steady-state optimization idea is proposed,in which reducing the energy consumption of coke-oven is taken as the optimization objective,the permitted open range of the butterfly valve and the technological requirements to gas collector pressure are used as the constraint conditions.Firstly,RBF neural network method based on linear regression and nearest neighbor-clustering algorithm is adopted to build the dynamitic set-point model of gas collector pressure,and the estimation set-point of gas collector pressure is obtained.Then,the target function for reducing energy consumption is build,and an improved particle swarm optimization algorithm is designed to find the steady-state optimization set-points of gas collector pressure in the target function.Actual application results show that the obtained optimization result using the proposed method can meet the requirements of the enterprise and good industrial application effect is achieved.
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2011年第4期769-774,共6页
Chinese Journal of Scientific Instrument
基金
湖南省高等学校科学研究项目(09C1020)资助
关键词
集气管压力
线性回归
RBF神经网络
粒子群优化
gas collector pressure
linear regression
RBF neural network
particle swarm optimization