Control Performance Assessment Based on Variance Decomposition for a Class of Nonlinear Cascade Systems

The minimum variance lower bound (MVLB) represents the best control performance in an output variance sense.For a class of nonlinear cascade systems,the traditional linear assessment method will cause an overlarge estimate on the system performance.In this work,the existence of MVLB of the block-oriented nonlinear cascade system is firstly proved,which just depends on the disturbances of several primary loop and secondary loop.And then,the system input and output data is utilized to reconstruct polynomial models for approximating describing the nonlinear cascade systems.By means of variance decomposition,Monte Carlo strategy is used to calculate the contribution of these disturbances items and determine MVLB estimate.Finally,it is shown from the comparison with linear assessment method that the proposed method is efficient and accurate for this class of nonlinear cascade systems.