时变迭合AR模型的参数估计

Parameter Estimation of Time Varying Mixed AR Model

首次提出了时变迭合ＡＲ模型，该模型在实际应用中具有广泛的应用价值．应用两步最小二乘法和限定记忆递推最小二乘法，给出了模型中时变参数的递推估计算法，该算法仅依靠量测数据即能自适应进行．仿真计算及应用结果表明：算法能够自适应地跟踪量测数据模型参数的变化，效果是令人满意的．

In systematic identification, the measure noise can be assumed as AR model in general. As the measure noise in measured data can't be observed, the parameter estimation in AR model is difficult. Especially, the parameters in AR model are dynamical with time in actual. In this paper, the time varying mixed AR model is first presented and is used to solve the above problem. Applying two step LS method and restricted remember LS method, an adaptive algorithm is proposed to estimate the time varying parameters only depending on measured data. Simulation and actual results show that the algorithm can automatically track the time varying characteristics of data. Applying this algorithm, we can also provide the statistical property needed by Kalman filtering in time.