基于信息滤波的极大似然递推辨识方法
发布时间:2018-05-12 22:41
本文选题:极大似然辨识 + 信息滤波 ; 参考:《江南大学》2017年博士论文
【摘要】:由于工业生产过程受到多种因素的干扰,使得有色噪声干扰系统的辨识变得更加困难.论文通过极大化似然函数,结合信息滤波技术研究一类有色噪声线性系统和非线性系统的参数辨识问题,选题具有理论意义和应用前景.取得了如下的成果.(1)针对标量方程误差ARMA系统,提出了基于信息滤波的极大似然增广梯度算法;为了加快梯度算法的收敛速度,引进新息,提出了基于信息滤波的极大似然多新息增广梯度算法.为了提高参数估计精度,推导了基于信息滤波的极大似然增广最小二乘算法.进一步,将提出的算法推广到多变量方程误差ARMA系统的参数估计中.(2)针对类多变量受控自回归ARMA系统,由于多变量系统的变量多、维数大,将系统分解成m(m是系统输出的维数)个子系统,再利用信息滤波对每个子系统的输入输出数据进行滤波,提出了子系统的信息滤波极大似然增广梯度算法,并与极大似然广义增广梯度算法进行比较,减少了有色噪声对参数估计的影响,提高了参数辨识精度.(3)针对多输入非线性Box-Jenkins系统,为了解决其非线性及有色噪声干扰的困难,利用分解技术,研究了基于分解的极大似然广义增广最小二乘辨识算法;针对有色噪声的干扰,选取适当的滤波器,对系统的输入输出数据进行预处理,再结合极大似然方法,提出了基于信息滤波技术的极大似然广义增广梯度辨识算法,减少了参数估计误差.论文中,对所提出的辨识算法都结合了数值算例,进行了仿真试验,通过仿真例子验证了算法的有效性。
[Abstract]:Due to the interference of many factors in the industrial production process, it is more difficult to identify the colored noise jamming system. In this paper, the problem of parameter identification for a class of linear and nonlinear systems with colored noise is studied by means of maximum likelihood function and information filtering technique. The following results are obtained: (1) aiming at the scalar equation error ARMA system, a maximum likelihood augmented gradient algorithm based on information filtering is proposed, and in order to speed up the convergence of the gradient algorithm, new information is introduced. A maximum likelihood multi-innovation augmented gradient algorithm based on information filtering is proposed. In order to improve the accuracy of parameter estimation, a maximum likelihood augmented least square algorithm based on information filtering is derived. Furthermore, the proposed algorithm is extended to the parameter estimation of multivariable equation error ARMA system. (2) for multivariable controlled autoregressive ARMA system, because the multivariable system has many variables and large dimension, The system is decomposed into three subsystems, which are the dimension of the system output. Then the input and output data of each subsystem are filtered by information filtering, and the information filtering maximum likelihood augmentation gradient algorithm of the subsystem is proposed. Compared with the maximum likelihood generalized augmented gradient algorithm, the influence of colored noise on parameter estimation is reduced, and the accuracy of parameter identification is improved. For the multi-input nonlinear Box-Jenkins system, it is difficult to solve the nonlinear and colored noise disturbance. Using decomposition technology, the maximum likelihood generalized augmented least square identification algorithm based on decomposition is studied, and the appropriate filter is selected to pre-process the input and output data of the system, and then the maximum likelihood method is combined with the maximum likelihood method. A maximum likelihood generalized augmented gradient identification algorithm based on information filtering technique is proposed to reduce the error of parameter estimation. In this paper, the proposed identification algorithms are combined with numerical examples, and the effectiveness of the algorithm is verified by simulation examples.
【学位授予单位】:江南大学
【学位级别】:博士
【学位授予年份】:2017
【分类号】:N945.14
【参考文献】
中国期刊全文数据库 前10条
1 丁锋;徐玲;刘喜梅;;信号建模(3):多频信号模型的递推参数估计[J];青岛科技大学学报(自然科学版);2017年03期
2 丁锋;徐玲;刘喜梅;;信号建模(2):双频率信号[J];青岛科技大学学报(自然科学版);2017年02期
3 丁锋;徐玲;刘喜梅;;信号建模(1):单频率信号[J];青岛科技大学学报(自然科学版);2017年01期
4 孙海洋;张利;;无人机跟踪场景下的粒子滤波算法的改进[J];计算机仿真;2017年02期
5 陈亚静;蔡如华;吴孙勇;桂丛楠;;基于粒子滤波的股价预测方法[J];统计与决策;2017年03期
6 肖震宇;于舒春;刘爽;于晓洋;;基于改进维纳滤波的运动模糊仪表图像恢复算法[J];电测与仪表;2017年02期
7 丁锋;汪菲菲;;损失数据线性参数系统的递推最小二乘辨识方法[J];控制与决策;2016年12期
8 蔡磊;王艳;纪志成;;基于折息递推辨识算法的数控机床能效预测[J];系统仿真学报;2016年08期
9 丁锋;;系统辨识算法的复杂性、收敛性及计算效率研究[J];控制与决策;2016年10期
10 刘艳君;丁锋;;多变量系统的耦合梯度辨识算法与性能分析[J];控制与决策;2016年08期
中国博士学位论文全文数据库 前1条
1 李俊红;极大似然辨识方法的研究[D];江南大学;2013年
,本文编号:1880460
本文链接:https://www.wllwen.com/shoufeilunwen/jckxbs/1880460.html