基于多新息算法的阶次未知的Wiener系统参数估计

Parameter estimation of unknown order Wiener system based on multi-innovation algorithm

针对阶次未知Wiener系统辨识模型过参数问题和最小二乘精度低的问题,提出一种行列式比确定阶次和基于分解技术的多新息最小二乘估计方案。首先,利用系统数据构造数据矩阵,利用行列式比方法估计系统的阶次;然后,利用分解技术将线性模块代入到非线性模块的特定项中,建立线性参数和非线性参数相互分离的估计模型,减少算法的计算量;其次,设计参考模型处理估计模型中存在的未知的内部变量,使内部变量转化为间接可测的变量;最后,采用一定的新息长度修改标量新息为多新息,提高估计算法的性能。通过分析不同噪声和不同新息长度对所提出算法的影响,验证该算法的鲁棒性和有效性。仿真结果表明,所提出的估计方案在估计精度和收敛速率上都优于递归最小二乘方法。

Aiming at the over-parameter problem of identification model forthe unknown order Wiener system and the low accuracy of the least-squares method,a determinant ratio method and the multi-innovation least squares approach based on decomposition technique were proposed.Firstly,data matrix was constructed from system data,and the order of the system was estimated by determinant ratio method.Secondly,the linear model was substituted into the special term of the nonlinear model by using the decomposition technique,and an estimation model of parameter separation of linear and nonlinear model was established,which reduced the computation burden of the algorithm.Then,a reference model was developed to handle the internal signal,which transformed the unknown internal signal into the indirectly measurable signal.Finally,scalar innovation was revised to multi-innovation through the usage of the certain length,which improved the performance of scheme.Furthermore,the influence of different noise and different innovation length on the proposed algorithm were analyzed.The simulation results showed that the proposed estimation scheme outperformed the recursive least squares method in estimation accuracy and convergence rate.