一种新型混合模拟退火算法及其应用  

A new hybrid simulated annealing algorithm and its application

作　　者：陈惟岐[1] 于镝[2] 

机构地区：[1]大庆石油学院秦皇岛分院,河北秦皇岛066004 [2]大庆石油学院电气信息工程学院,黑龙江大庆163318

出　　处：《大庆石油学院学报》2007年第4期84-87,共4页Journal of Daqing Petroleum Institute

基　　金：黑龙江自然科学基金资助课题(TF2005-26)

摘　　要：为解决非线性系统辨识和预测问题,以多层前向网络为模型框架,采用带自适应冷却进度表的模拟退火算法与Powell算法构成新型混合算法,训练网络的权值.冷却进度表中的主要参数是模拟退火算法的控制参数(伪温度)的初值及其衰减函数.把整个迭代过程划分为若干阶段.在每个阶段结束时,依据网络模型误差自适应地修正下阶段的回火温度、伪温度的衰减函数中的参数和迭代步长初值.该混合算法的全局和局部搜索能力强,网络辨识精度高.应用表明了该方案的有效性.In this paper,multilayer feedforward neural networks model is framed to solve the problem in nonlinear system identification and prediction.To train the weights of NN,it proposed a new hybrid algorithm consisting of Powell algorithm and simulated annealing algorithm with adaptive cooling schedule.The main parameters in the cooling schedule are initial value and the attenuation function of control parameter(fictive temperature) in the algorithm.The whole iteration is divided into several stages.When each stage ends,according to NN model errors,the next stage's tempering temperature,the parameter of attenuation function of fictive temperature,and initial value of iterative step-length are adaptively modified.The above hybrid algorithm has strongly global and local search property and identification precision is remarkably raised.The validity of the scheme is proved by its applications.

关 键 词：神经网络 模拟退火算法 冷却进度表 POWELL算法 辨识 非线性系统 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]