多变量系统的变时滞无模型滤波PID控制

Variable-time-delay and model-free filtering PID control in multivariable system

多变量紧格式动态线性化泛模型仅适用于常值干扰和慢变化干扰情形。其结构自适应功能只对系统的输出阶数和输入阶数有效,对系统的时滞无效,同时其伪梯度矩阵参数不唯一,要求控制输入的变化量不能为零。为此,提出一种适用于快变化干扰和随机干扰的多变量紧格式动态线性化泛模型,采用多变量解耦增量型滤波PID控制,基于可克服算法病态的非线性递推最小二乘算法对PID控制参数寻优,给出多变量系统的在线修正参数的变时滞无模型滤波PID控制算法。结果表明,算法具有在线修正参数性能和无模型自适应控制功能,以及优良的控制品质。

This paper is intended to address an inherent limitation which plagues the universal model of multivariable compact form dynamic linearization in that it is only applicable to constant interference and slow state interference, and its structural adaptation function works only for the output order and input order of the system rather than for the time delay, and what is more, its pseudo-gradient matrix parameter is not unique and the change of control input cannot be zero. The paper proposes the universal model of multivariable compact form dynamic linearization, which is better able to work for fast state interference and random disturbance, by separating fast state interference and random disturbance from system model to overcome the above problems by adding auxiliary vector and time-varying time-delay to the universal model. The paper highlights a variable-time-delay and model-free filtering PID control algorithm with on-line modifying parameters is proposed in multivariable system based on nonlinear recursive least squares algorithm which can overcome algorithm turning ill-posed and using multivariable decoupling incremental filter PID control optimize PID control parameters. Simulations show that the proposed algorithm could provide on-line modifying parameter performance and model-free adaptive control function, and therefore a better control quality.