摘要
以方程残差作为目标函数 ,运用基因遗传算法同时辨识受控系统的反馈时滞和物理参数。由于遗传算法具备良好的全局特性 ,且不要求目标函数光滑 ,该方法既适于线性系统的辨识也适于非线性系统的时滞辨识。方法中引入几种交叉和变异算子同时操作 ,大大提高了辨识效率和辨识精度。当需辨识参数过多时 ,可先用开环实验辨识部分系统参数 ,再用闭环实验辨识时滞和剩余的系统参数。
The experimental modeling of delayed dynamic systems is an open problem.This paper presents an approach based on the genetic algorithms to estimate the system parameters and time delays in the feedback simultaneously through the minimization of residual error of system equation.Several operations such as crossover and mutation are combined so as to improve the convergence rate in the process of parametric estimation.The approach is illustrated by two examples of numerical simulation.Due to the global optimum in a parallel,random and blind way,the approach works for both linear and nonlinear systems with delays.In the case of many system parameters to be estimated,it is suggested to estimate a part of system parameters through an open loop experiment and then the time delays and the remaining parameters through a closed loop experiment.
出处
《振动工程学报》
EI
CSCD
北大核心
2001年第3期314-318,共5页
Journal of Vibration Engineering
基金
国家杰出青年学基金 (编号 :5 96 2 5 5 11)
国家自然科学基金 (编号 :19972 0 5 5 )资助项目
关键词
非线性振动
时滞
主动控制
基因遗传算法
系统辨识
nonlinear vibration
time delay
active control
genetic algorithms
system identification
