混沌混合粒子群算法应用于优化酯化反应神经网络模型  

Chaotic hybrid particle swarm optimization algorithm and i ts application in optimization of ANN model of esterification

作　　者：张立庆[1] 程志刚[2] 吴晓华[3] 李小林[1] 盛煜翔[1] 

机构地区：[1]浙江科技学院生物与化学工程学院,浙江杭州310012 [2]浙江科技学院信息与电子工程学院,浙江杭州310012 [3]浙江大学材料与化学工程学院,浙江杭州310027

出　　处：《计算机与应用化学》2007年第8期1073-1077,共5页Computers and Applied Chemistry

基　　金：浙江省自然科学基金项目(Y404082);浙江省教育厅重点科研计划项目(20030836)

摘　　要：利用混沌映射的随机性和遍历性,将其引入粒子群优化算法,以提高算法的全局寻优能力,同时引入优进策略,以改善其局部寻优效率,在此基础上构建了混沌混合粒子群优化算法(CHPSO)。高维复杂函数的仿真优化试验表明,CHPSO全局寻优能力强、优化效率高。针对常规算法训练神经网络容易早熟收敛和陷入局部极值点的不足,采用CHPSO训练人工神经网络,由此构建CHPSO-NN模型,并应用于乙酸己酯催化酯化反应条件的建模,与BP-NN相比,其预测能力和稳健性都有较大提高,效果良好,与传统方法相比有明显的优越性。a new algorithm, which is named as chaotic hybrid particle swarm optimization algorithm （CHPSO） , is proposed. CHPSO integrates chaotic mechanism for its ergodicity, stochastic property, and regularity, which enhance the global exploitation. And also integrates the eugenic strategy to improve the local exploration. Simulation results on high dimensional complex functions shows that CHPSO has powerful global optimization ability and high optimization efficiency. Then CHPSO was proposed to training Neural Networks, named CHPSO-ANN, to establish reaction model for esterification of hexyl acetate, the predict ability and stability of results have a quite increase compared with the traditional BP-ANN method.

关 键 词：混沌 优进 粒子群优化 神经网络 酯化 

分 类 号：O434.2[机械工程—光学工程]