摘要
针对一致关联度算法不具有普遍性和动态改变惯性权的自适应粒子群算法(DCW)不易跳出局部收敛能力的缺陷,本文提出了完全关联度算法和自适应变异的动态粒子群优化算法。完全关联度算法主要用来选择软测量的辅助变量。在改进的粒子群优化算法中,除了采用动态惯性权重外,还引入了自适应学习因子和新的变异算子。为了构造一种性能较好的神经网络,采用改进的粒子群优化算法来优化神经网络所有的权值参数,并将提出的软测量建模方法预测延迟焦化的汽油干点,实验结果表明,与DCW算法优化神经网络(DCWNN)的建模方法相比,该算法不仅具有较好的泛化性能,而且具有较高的精度和良好的应用前景。
A novel perfect incidence degree algorithm and adaptive mutation dynamic particle swarm optimization algorithm are presented,to solve the problem that uniform incidence degree has not universality and particle swarm algorithm with dynamically changing inertia weight（DCW） is is easy to fall into local optimization.Perfect incidence degree algorithm is used to seleet seeondary variables of soft sensor.In improved particle swarm optimization algorithm,besides that it makes use of adaptive inertia weight,adaptive term coefficients and new mutation operator are introduced to the algorithm.In order to construct a better performance of neural networks,improved particle swarm optimization algorithm is used to optimize all parameters of neural networks.The proposed method has been applied to predict the gasoline end point in delayed coking.The experimental results show that the algorithm not only has better generalization performance but also has higher precision and good prospect of application compared to soft sensor modeling method of DCW optimization neural networks（DCWNN） algorithms.
出处
《微计算机信息》
2011年第1期285-287,共3页
Control & Automation
关键词
软测量
完全关联度
粒子群优化
神经网络
汽油干点
soft sensor
perfect incidence degree
particle swarm optimization
neural network
gasoline end point
作者简介
贺晓春（1981－），男，（汉族），陕西宝鸡人，网络工程师，助教．西南交通大学工程硕士在读．主要研究方向：计算机应用研究．通讯地址：（628017四川省广元市利州区奔月路四川信息职业技术学院）贺晓春
