基于Alopex的进化算法及其在乙炔加氢反应器建模中的应用  

A New Evolutionary Optimization Algorithms and Its Application in Soft Sensor Modeling

作　　者：李绍军[1] 张小广[1] 

机构地区：[1]华东理工大学自动化研究所,上海200237

出　　处：《化工自动化及仪表》2008年第3期62-66,共5页Control and Instruments in Chemical Industry

基　　金：上海市自然科学基金资助项目(06ZR14027);上海市重点学科建设项目(B504)

摘　　要：分析进化优化算法的共性,参考Alopex优化思想提出一种新的进化优化算法。给出该算法的具体步骤,将该算法与标准粒子群算法、遗传算法和差分算法在性能上做了仿真对比,并将其应用于乙炔加氢反应器出口乙炔浓度软测量的建模中。这种新的优化算法具有全局收敛能力,并具有较快的收敛速度。对典型函数的测试和基于神经网络的软测量建模表明:新算法具有较强的全局搜索能力,特别是对易于陷入局部最优的多峰函数能够有效地避免早熟收敛问题。The generality of evolutionary optimization algorithm was analyzed. A new evolutionary algorithm was proposed based on Alopex and the steps of the algorithm were given. Compared the effectiveness of the new algorithm with PSO, GA and DE through simulation. The algorithm was used in the soft sensor modeling of acetylene hydrogenation reactor outlet concentration. The new algorithm has the ability of global convergence, and the quick convergence rate. According to the results of function test and the soft sensor modeling based on neural network, the proposed algo- rithm is superior to original evolutionary algorithms, especially multi-apices function which easily fall into local optimum.

关 键 词：进化算法 ALOPEX算法 软测量 神经网络 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]