遗忘因子选取对于时变参数识别结果的影响

本文在已知后验知识的基础上，提出了以极小化待识别参数标准差为目标函数的最佳遗忘因子选择方案，考察了遗忘因子选取对时变参数识别结果的影响。数字仿真结果表明，对于各种不同算法（ＬＳ，ＤＬＳ，ＩＶ），选用最佳遗忘因子后，相应于标准差达极小的物理参数的识别结果均优于以往凭经验所选择的遗忘因子的识别结果。依本文方法而进行的模拟计算，可以为各种不同变化类型的时变参数的识别提供一种选择最佳遗忘因子的理论依据

EFFECT OF FORGETTING FACTORS ON ESTIMATED RESULTS OF TIME VARYING PARAMETERS

There have been no methods of selecting the ideal values of the forgetting factors so far.In this paper,an optimum scheme for determining this parameter is presented on the basis of a knowledge after effect,where the normalised error of the estimated parametrs is taken as an objective function. The effect of forgetting factors on estimated results of time varying parameters is examined. The simulated results show that as the optimum forgetting factor is determined, the estimated results of the physical parameter corresponding to the minimum noumalised error are better than those obtained by using experience to determine the forgetting factors for three time domain system identification algorithms:Least Squares,Double Least Squares and Instrumental Variables.The simulated calculations according to this method can provide a theoretical basis for determining the optimum forgetting factor to the identification of time varying parameters of different varying types.