应用最优化方法辨识指数自回归模型

本文研究应用非线性最优化原理辨识指数自回归模型的方法，以Ｎｅｗｔｏｎ一维搜索和线性回归法为基础，提出一种改进的坐标轮换算法，交替地估计模型的线性和非线性参数。实例表明，该算法对指数自回归模型具有良好的辨识效果。

Identification of EAR Model Based on Optimization Theory

xponential autoregressive (Ear) model is a nonlinear time series model which is introduced to reproduce certain features of nonlinear random vibration. In this paper, we present an approach for identifying EAR model. The proPOsed approach is based upon optimization theory. The parameter estimation algorithm is formulated as an iterative procedure. In each iteration step, the linear and nonlinear parameters of EAR model are estimated alternatively. The linear riarameters are estimated by recursive least squares algorithm, and the nonlinear parameter is estimated by modified Newton's algorithm. The model order is determined by applying AIC criterion. Simulation examples are given to demonstrate the efficiency of the proposed method.