摘要
目前的辨识方法一般需要在系统输入端加入激励信号 ,而且多输入多输出系统的在线辨识仍很困难。本文提出一种基于牛顿迭代法的多输入、多输出对象模型迭代辨识方法 ,模型参数更新的依据是使模型预测输出与全部采样时刻的对象实际输出之间的均方差递减 ,直到收敛。这种基于全局数据迭代的辨识方法可进行闭环辨识 ,无需外加激励信号 ,适用于多输入多输出对象的在线辨识。对一个两输入、两输出对象模型的仿真研究和某电厂 30 0MW机组负荷被控对象的计算结果表明 ,辨识效果令人满意。
Almost all of the normal identification methods need stimulating signal added into the input of the system.And their behavior in online identification of multi-input multi-output system is not good.The paper developed an iteration identification method for multi-input multi-ouput system based on the Newton iteration method.This identification method based on global data iteration.It updates the model parameters by minimizing the mean square deviation between the predicted output by the model and the real output of the object during the whole sampling time until the iteration converges.This method can deal with the closed loop identification and be free from stimulating signal.The calculating results for a two-input two-output model and a 300MW fossil-fired power generation unit showed good performance of the strategy.
出处
《自动化与仪器仪表》
2003年第4期36-38,50,共4页
Automation & Instrumentation
基金
国家自然科学其金 (50 1 760 56)
高等学校优秀青年教师教学和科研奖励计划 (1 999年度 )资助
