基于主元回归方法的多变量控制系统性能评估

Multivariable Control System Performance Assessment Based on Principal Component Regression Method

针对最小方差评价方法利用不切实际和不期望的指标进行性能评价所造成的性能评估缺陷,提出基于主元回归(PCR)的性能评价方法。该方法通过对输入输出数据进行建模可以得到比较符合实际的理论方差,以此理论方差与实际输出方差求出的控制系统性能指标,更能客观地反映控制系统的真实性能,且主元回归性能评价方法在性能计算中不需求解关联矩阵问题,极大地简化了求解的复杂度。最后通过仿真实例验证了主元回归性能评价方法更能客观反映控制系统真实性能和计算相对简易的优点。

Considering the performance assessment defects caused by the minimum variance evaluation method of using impractical index and the index not expected,the performance evaluation method based on principal component regression（PCR） was proposed to get a relatively realistic theory variance by the modeling based on input and output data;the performance index of control system can be computed by the theory variance and actual output variance to objectively reflect the real performance of control system,and the performance evaluation method based on principal component regression in performance calculation which demanding no associated matrix can vastly simplify the calculation.The simulation results show that this method can reflect actual performance of the contorl system and can be calculated easily.