模型辨识新方法及应用

A New Approach for Model Identification and Its Application

本文提出了一种有效的模型辨识新方法．为了提高数值稳定性和计算效率，本文给出了一种递阶最大信息量（ＡＩＣ）新判据和参数估计新方法，使单输出系统的极大似然方法及模型辨识的优选判据计算量成倍减少．通过分析标量系统的ＡＩＣ标准，本文进一步导出了多输出情况下的ＡＩＣ标准，大大提高了计算效率．

A new efficient method for nonlinear model identification is presented in this paper. To get high numerical stability and computational efficiency,a new AIC criterion and recursive parameter estimation methods are developed which make both the maximum likelihood method and the optimal criterion for model choice of SISO system very simple.For MIMO system,a new AIC criterion is given and the computation formulas of the cost function is simplified.The results of aerodynamic model identification show that the new method can give accurate model structures of unknown system.