时变遗忘因子的子空间辨识及预测控制器设计

Predictive Controller Design Based on Subspace Identification with Varying Forgetting Factor

摘要:

针对线性时变多变量系统,在存在不可测干扰及系统动态特性变化较大的情况下,不需要已知系统先验结构信息,不需要辨识出系统参数矩阵,提出一种完全数据驱动的具有变遗忘因子子空间辨识的预测控制器设计方法.预测控制是一种基于模型的控制方法,为了更好地建立被控系统模型,在已有的在线辨识基础上,根据实测输出值与预测输出值的误差构造变遗忘因子,以调整采集数据的权重,提高辨识灵敏度和控制效果.最后通过实例仿真验证算法的有效性.

Abstract:

In the existence of unpredictable disturbance and large changes in dynamic characteristics,a complete datadriven method based on subspace identification with variable forgetting factor is proposed for LTV multivariable system,without any priori structural information and identification of system parameters matrix.Predictive control is a modelbased control method.In order to establish better system model,the variable forgetting factor is structured by error of the real and predictive output value based on on-line identification.Thus the weight of the collection data is adjusted and identification sensitivity and control effect is improved.Finally,a simulation example is given to demonstrate the efficiency of the proposed algorithm.