含缺失数据的EIV系统辨识  

Identification of EIV Systems with Missing Data

作　　者：黄佳成 谢莉[1] HUANG Jia-cheng;XIE Li(School of Internet of Things Engineering,Jiangnan University,Wuxi 214122,China)

机构地区：[1]江南大学物联网工程学院,江苏无锡易214122

出　　处：《控制工程》2023年第1期32-38,共7页Control Engineering of China

基　　金：国家自然科学基金资助项目(61403166,61773181);江苏省自然科学基金资助项目(BK20140164)。

摘　　要：针对含缺失数据的变量带误差(EIV)系统,直接利用协方差匹配(CM)算法进行辨识的精度有限,为此提出一种协方差匹配迭代(covariance matching based iterative,CMI)算法。首先基于不完整数据集,利用CM算法获得模型参数的初始估计,然后采用交互估计理论,利用获得的参数计算缺失输出数据的估计,重构得到完整的数据集后再进一步利用CM算法更新参数估计。两者执行了递阶计算过程,通过迭代辨识逐步提高参数估计精度。仿真结果表明,CMI算法的参数估计误差在输出数据缺失率达到60%时仍然能够保持在2%以下,且随输入端和输出端噪信比的变化速率仅为CM算法的16.8%和10.8%,验证了所提算法具有较高的辨识精度和良好的鲁棒性。For errors-in-variables(EIV)systems with missing data,the identification accuracy was limited as the covariance matching(CM)algorithm was used directly,thus a covariance matching based iterative(CMI)algorithm was proposed.Firstly,the initial estimates of model parameters were obtained by using the CM algorithm based on incomplete data sets.Then,the obtained parameters were applied to calculate the estimates of the missing outputs with the interactive estimation theory,and the CM algorithm was further utilized to update the parameter estimates based on the reconstructed complete data set.The above two steps carried out a hierarchical calculation procedure,and the parameter estimation accuracy could be gradually improved through the iterative identification.Simulation results show that the parameter estimation error of CMI algorithm can remain below 2%when the missing rate of output data reaches 60%,and the change rate of parameter estimation error with noise to signal ratio for input and output is only 16.8%and 10.8%,respectively of the CM algorithm,which verifies that the proposed algorithm has high identification accuracy and great robustness.

关 键 词：变量带误差系统 协方差匹配方法 迭代辨识 缺失数据 参数估计 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]