基于多新息最小二乘算法的非线性系统辨识

Nonlinear System Identification Based on Multi-Innovation Least Squares Algorithm

针对最小二乘算法辨识性能较差问题,将最小二乘算法中的单新息通过利用p组数据拓展到多新息向量,提出了多新息最小二乘算法。与最小二乘相比,所提出的算法不仅利用了当前的系统信息,而且利用了过去的系统信息,进一步提高了参数辨识的精度和收敛速度。在所提出的算法中,为了减少冗余的参数辨识和算法计算量,利用关键性分离技术构造整体辨识模型。设计了辅助模型来替代系统中未知的中间变量,提高了参数估计的精度。对比仿真结果表明,所提出的算法具有比递归最小二乘算法更高的辨识精度和收敛速度。

In view of the poor identification performance of the least square algorithm,the multi-innovation least squares algorithm is proposed by expanding the single innovation to multi-innovation.Compared with least squares algorithm,the proposed algorithm can use not only the current system information but also the past system information,which improves the convergence rate and identification accuracy.In order to decrease the redundant parameter estimation and computational burden,the key separation technology is used to establish the identification model.An auxiliary model is designed to replace the unknown intermediate variables in the system,which improves the accuracy of parameter estimation.The comparative simulation shows that the proposed algorithm has higher identification accuracy and convergence speed than the recursive least squares algorithm.