论文部分内容阅读
将参数检测技术和辨识方法相结合、系统结构在线辨识和参数跟踪相结合,基于U—D分解技术,提出一种时变系统结构确定和参数估计的最小二乘辨识新算法(MUDI).该算法不仅可实现系统阶次和参数的同时估计,而且通过对损失函数的实时监测,实现协方差阵的自适应调整,使辨识算法收敛速度快,对时变系统阶次和参数变化均有很强的跟踪能力.此外,由于采用U—D分解技术,与递推最小二乘法(RLS)等相比,本文的算法不仅具有很好的数值稳定性和快速收敛性,而且计算量明显小于RLS,仿真计算结果表明本文算法的有效性和优越性.
Based on U-D decomposition technique, a new least square identification algorithm (MUDI) for structure determination and parameter estimation of time-varying system is proposed based on the combination of parameter detection technology and identification method, on-line identification of system structure and parameter tracking. The algorithm can not only estimate the system order and parameters simultaneously, but also realize the adaptive adjustment of the covariance matrix through the real-time monitoring of the loss function, so that the convergence speed of the identification algorithm is fast and the order and parameter changes of the time-varying system are Strong tracking ability. In addition, compared with Recursive Least Squares (RLS) and so on, U-D decomposition method not only has good numerical stability and fast convergence, but also has less computational complexity than RLS. Simulation results show that The effectiveness and superiority of this algorithm.