连续Hammerstein模型直接辨识方法研究

发布时间：2018-04-11 18:06

本文选题：参数估计 + 连续Hammerstein模型　；参考：《长沙理工大学》2011年硕士论文

【摘要】：课题以复杂机械系统动力学建模与设计为工业应用背景,在国家自然科学基金项目《连续非线性动力学系统参数模型小波调制直接辨识》(项目编号:50875028)资助下,拟定以带测量噪声情况下连续Hammerstein模型直接辨识为研究内容。主要进行了以下几个方面的研究: 1.建立了无测量噪声干扰的连续Hammerstein模型的调制最小二乘直接辨识算法。介绍了非线性参数分离方法,用分离方法分解重整模型参数向量得到的乘积项,最终得到连续Hammerstein模型的估计参数。 2.建立了有测量噪声干扰情况的连续Hammerstein模型的调制最小二乘直接辨识算法。研究其噪声的调制最小二乘特性,知其参数估计是有偏参数估计,利用广义噪声模型处理调制噪声,提出调制广义最小二乘算法获得了系统参数的无偏估计。 3.针对高噪声情况下调制广义最小二乘参数估计出现失真的情况,提出时窗小波降噪算法对带高噪声的数据进行在线降噪,然后利用降噪数据估计模型参数。研究了时窗小波降噪在连续线性动力学模型和连续Hammerstein模型中的应用,由于降噪数据也带有低噪声,故分别提出时窗小波降噪调制辅助变量法和时窗小波降噪调制广义最小二乘算法来获得相应模型参数的高精度的无偏估计。 4.对冷轧平整机HAGC压力闭环系统建立连续线性动力学参数模型和连续Hammerstein模型。将提出的算法用来辨识相应的模型参数,得到较好的辨识结果。应用实例说明了算法的有效性和实用性。研究表明:本文研究的辨识算法对测量噪声干扰下估计连续Hammerstein模型参数有较好的抗干扰能力,可适应大型工业系统的模型辨识。
[Abstract]:Under the background of complex mechanical system dynamic modeling and design, supported by the National Natural Science Foundation of China, "Direct Identification of Parameter Model of continuous nonlinear dynamic system by Wavelet Modulation" (item No.: 50875028),Direct identification of continuous Hammerstein model with measurement noise is proposed.Mainly carried out the following aspects of research:1.A modulation least square direct identification algorithm for continuous Hammerstein model without measured noise interference is established.The nonlinear parameter separation method is introduced. The product term obtained from the parameter vector of the reforming model is decomposed by the separation method, and the estimated parameters of the continuous Hammerstein model are obtained.2.A modulation least square direct identification algorithm for continuous Hammerstein model with measured noise interference is established.The modulation least square characteristic of the noise is studied, and the parameter estimation is known as biased parameter estimation. The generalized noise model is used to deal with the modulation noise, and the modulation generalized least squares algorithm is proposed to obtain the unbiased estimation of the system parameters.3.Aiming at the distortion of modulated generalized least-squares parameter estimation under high noise, a time-window wavelet denoising algorithm is proposed to reduce the noise of data with high noise on line, and then the model parameters are estimated by denoising data.The application of time-window wavelet denoising in continuous linear dynamic model and continuous Hammerstein model is studied.Therefore, the time window wavelet denoising modulation auxiliary variable method and the time window wavelet denoising modulation generalized least square algorithm are proposed to obtain the high accuracy unbiased estimation of the corresponding model parameters.4.A continuous linear dynamic parameter model and a continuous Hammerstein model are established for the HAGC pressure closed loop system of cold rolling mill.The proposed algorithm is used to identify the corresponding model parameters, and better identification results are obtained.An example is given to illustrate the effectiveness and practicability of the algorithm.
【学位授予单位】：长沙理工大学
【学位级别】：硕士
【学位授予年份】：2011
【分类号】：TH113

【参考文献】