AR模型参数自适应估计方法研究及应用

为了提高AR模型参数估计的精度和收敛速度,提出基于卡尔曼(Kalm an)滤波原理的AR模型参数估计方法.把AR模型参数向量作为状态向量,利用随机游动模型建立系统的状态方程,由观测数据建立系统的观测方程,应用卡尔曼(Kalm an)滤波原理求得参数向量的最小方差估计.将该方法应用于船舶运动实时建模预报中.仿真研究表明本文提出的基于Kalm an滤波算法的AR模型在预报精度以及收敛速度方面均优于基于递推最小二乘法(RLS)和最小均方(LMS)算法,该算法具有较强的鲁棒性,降低了实时在线预报时通信故障所引起的数据缺失对预报精度的影响.该方法在理论和工程应用方面具有重要的意义.

Research and Application On AR Model Parameters Adaptive estimation

In order to improve parameter estimation precision and convergence rate about autoregressive (AR) model, It was proposed that a new approach of AR model parameter estimation based on the kalman filtering principle. And it was applied in the ship motion real-time modeling and forecasts. Compared with recursive least-square(RLS) algorithm and least mean square(LMS) algorithm, the simulation results indicates the new parameter estimates approach based on the kalman filtering principle is better than RLS algorithm and LMS algorithm in the forecasting precision and convergence rate. The new approach has strong robustness, and reduced the influence of forecasting precision as result of the data flaw when correspondence breakdown occurred in the real-time online forecast. So the approach has vital significance in the theory and the project application aspect.