Experiments with Identification of Continuous Time Models
Experiments with Identification of Continuous Time Models
复制标题
连续时间模型辨识实验
DOI：
--
复制
发表时间：
2009
期刊：
影响因子：
--
通讯作者：
L. Ljung
中科院分区：
文献类型：
--
作者：
L. Ljung
从采样数据中识别时间连续模型是一个长期的讨论话题，已经提出了许多方法。极大似然法在理论上和渐近上都上级其它方法。然而，它可能遭受在快速采样的数值不准确，它也需要可靠的初始参数值。多年来，已开发出许多有效和有用的最大似然法替代方法。其中最重要的是状态变量滤波器，结合工具变量方法，包括简化的改进IV方法。在这方面的贡献，我们进行谦逊的数值实验来评论这些方法，他们的共同利益。
Identification of time-continuous models from sampled data is a long standing topic of discussion, and many approaches have been suggested. The Maximum Likelihood method is asymptotically and theoretically superior to other methods. However, it may suffer from numerical inaccuracies at fast sampling and it also requires reliable initial parameter values. A number of efficient and useful alternatives to the maximum-likelihood method have been developed over the years. The most important of these are State-Variable filters, combined with Instrumental Variable methods, including the simplified refined IV method. In this contribution we perform unpretentious numerical experiments to comment on these methods, and their mutual benefits.