 1. 大连理工大学 控制科学与工程学院,辽宁 大连 116024;2. 新疆大学 电气工程学院,乌鲁木齐 830047

E-mail: 1195201627@qq.com.

TP13

Multi-model switching identification for non-uniformly sampled systems based on Gaussian mixture model clustering
Author:
Affiliation:

1. School of Control Science and Control Engineering,Dalian University of Technology,Dalian 116024,China;2. School of Electrical Engineering,Xinjiang University,Urumqi 830047,China

Fund Project:

• 摘要
• |
• 图/表
• |
• 访问统计
• |
• 参考文献
• |
• 相似文献
• |
• 引证文献
• |
• 资源附件
• |
• 文章评论
摘要:

从概率统计方法出发,提出一种基于高斯混合模型聚类与递推最小二乘算法的非均匀采样系统的多模型建模方法.首先,采用高斯混合模型作为调度函数,使用最大期望(EM)算法迭代更新估计高斯混合模型中参数,从而通过每个子系统的高斯概率密度函数计算和比较来确定子系统的激活情况; 其次,采用递推最小二乘算法估计局部子系统参数;然后,使用鞅收敛定理对所提出的算法性能进行分析; 最后,通过非均匀采样系统的多模型建模来证明所提出方法的有效性.

Abstract:

Based on the probabilistic method, the multi-model modeling method of non-uniformly sampled systems based on the Gaussian mixture model clustering and recursive least square algorithm is proposed,. Firstly, the Gaussian mixture model is used as the scheduling function, and the Expectation-Maximization(EM) algorithm is used to iteratively update and estimate the parameters of the Gaussian mixture model, so that the activation of each subsystem can be determined by calculating and comparing the Gaussian probability density function of each subsystem. Secondly, the recursive least square algorithm is used to estimate the parameters of the local subsystem. Thirdly, the martingale convergence theorem is used to analyze the performance of the proposed algorithm. Finally, the effectiveness of the proposed method is proved by the multi-model modeling for the non-uniformly sampled system.

参考文献
相似文献
引证文献

• 点击次数:
• 下载次数:
• HTML阅读次数:
• 引用次数:
##### 历史
• 收稿日期:
• 最后修改日期:
• 录用日期:
• 在线发布日期: 2021-11-18
• 出版日期: 2021-12-20