应用混沌粒子群优化训练的BP神经网络预报高炉铁水含硅量

来源期刊：冶金分析2010年第5期

论文作者：徐生林 史燕 杨成忠

文章页码：9 - 13

关键词：混沌粒子群优化算法； BP神经网络； 铁水含硅量； 预报

Key words：chaos particle swarm optimization(CPSO)； BP neural network(BPNN)； silicon content in liquid iron； prediction

摘    要：应用混沌粒子群优化(Chaos Particle Swarm Opti mization,CPSO)算法训练BP神经网络(Back-Propagation Neural Network,BPNN)并对高炉铁水含硅量进行预报。针对BP神经网络收敛速度慢和易于陷入极小值的问题,在粒子群优化算法中引入混沌思想,提出混沌粒子群优化算法,改善了粒子群优化算法摆脱局部极值点的能力,提高了BP网络的计算精度和收敛速度。系统分别选用料速、透气性指数、炉顶温度、风温、风量、喷煤量、上一炉铁水硅含量作为BP神经网络的输入层神经元,中间层(隐含层)有13个结点(用经验公式确定),输出层有一个结点,为铁水硅含量。应用CPSO算法训练BP神经网络建立的铁水含硅量预报模型对江苏永钢炼铁一厂1号高炉铁水含硅量的实际数据进行网络学习和预报。结果表明,此模型预报命中率高达91.2%,证明了方法的有效性。

Abstract: A new approach to predict the silicon content in liquid iron was established based on back-propagation neural network (BPNN) trained by chaos particle swarm optimization (CPSO).An advanced particle swarm optimization algorithm based on chaos search was proposed to enhance the local searching ability.The calculation accuracy and convergence rate of the BP neural network were improved.In this system,the speed of feeding fuel,permeability index,temperature of blast top,air temperature,air flow,coal injection amount and the silicon content in liquid iron of former furnace were used as the input layer neurons of BP neural network.The middle layer (hidden layer) had 13 nodes (which were determined by empirical formula),and the output layer had one node (the silicon content in liquid iron).The established prediction model of silicon content in liquid iron based on BPNN trained by CPSO was used for network learning and predicting the real time data collected from No.1 BF of ironmaking No.1 factory in Jiangsu Yonggang Iron and Steel Group Co.,Ltd.The results showed that the hit rate of prediction model reached up to 91.2 %,indicating the validity of method.