Abstract: To evaluate the system performance timely,the Control Performance Assessment(CPA)problem of Single-Input-Single-Output(SISO)systems was investigated based on Minimum Variance Control(MVC)theory by considering the cases of zero and nonzero system set-points respectively.Aiming at the demand of existing linear regression and the defects of requiring excessive computing burden and invertible matrix calculation of data matrices,the recursive least-square technique was introduced to calculate the minimum variance benchmark.An improved algorithm based on minimum variance control was concluded for performance evaluation of SISO plants thus the shortages of the linear regression based method were avoided.The evaluation results of the improved algorithm and the existing linear regression algorithm were compared by simulation study to validate the effectiveness of this improved algorithm.

Key words: control system, minimum variance control, recursive least-square technique, performance assessment algorithm

摘要： 为了及时评价工业控制系统性能的优劣,维护和提高其运行效果,使生产过程和最终产品获得最佳效益,分别考虑了设定值为零和非零两种情况,对一类单输入单输出系统基于最小方差控制的性能评价算法进行了研究。针对已有的线性回归算法的需要,对数据矩阵求逆、计算量过大等可能导致算法崩溃的缺陷,采用递推最小二乘方法计算控制器的最小方差性能基准,从而避免了矩阵求逆等复杂运算,得到了一种改进的基于最小方差控制的性能评价算法,克服了上述线性回归算法的不足。通过仿真示例,将该改进算法和已有线性回归算法的评价结果进行对比分析,验证了改进算法的有效性。

关键词: 控制系统, 最小方差控制, 递推最小二乘法, 性能评价算法