Identifiability and Performance Analysis of Output Over-sampling Approach to Direct Closed-loop Identification
Identifiability and Performance Analysis of Output Over-sampling Approach to Direct Closed-loop Identification
复制标题
DOI：
10.9746/sicetr.45.339
复制
发表时间：
2009
期刊：
影响因子：
--
通讯作者：
Lianming Sun;A. Sano
中科院分区：
文献类型：
--
作者：
Lianming Sun;A. Sano
研究了基于输出过采样的闭环辨识算法。分析了连续随机噪声和对象输入、输出在过采样方法中的一些本质特性，并利用这些特性论证了过采样方法的可辨识性，评价了过采样方法的辨识性能。此外，还探讨了被控对象模型阶数的选择、参数估计的渐近方差和估计模型频率响应的渐近方差。结果表明，过采样方法在保证系统可辨识性的同时，大大提高了系统的闭环辨识性能。
Output over-sampling based closed-loop identification algorithm is investigated in this paper. Some instinct properties of the continuous stochastic noise and the plant input, output in the over-sampling approach are analyzed, and they are used to demonstrate the identifiability in the over-sampling approach and to evaluate its identification performance. Furthermore, the selection of plant model order, the asymptotic variance of estimated parameters and the asymptotic variance of frequency response of the estimated model are also explored. It shows that the over-sampling approach can guarantee the identifiability and improve the performance of closed-loop identification greatly.