模型结构分析方法在多变量预测控制中的应用

Model structure analysis in the application of multivariable predictive control

为了解决因在模型辨识阶段产生辨识误差而可能引起的稳态增益模型出现线性相关度增高的问题,研究了模型结构临界不稳定现象与多变量预测控制系统性能间的关系.首先,利用奇异值分解(SVD)和相对增益矩阵(RGA)等工具分析了稳态模型的线性相关度,从而给出了模型结构临界不稳定的判定方法.其次,给出了由建模误差引起的模型结构临界不稳定问题或过程控制系统设计与装置存在的不匹配问题时的控制系统改进方法,并通过仿真对比验证了模型结构对控制性能的影响及改进方法的有效性.

In order to solve the problem that identification error emerged in the model identification stage may cause the dependence of the steady state model higher, the relationship between the critical instability of the model structure and the performance of the multi-variable predictive control is investigated. First, the singular value decomposition （SVD） and the relative gain array （RGA） are used to analyze the linear dependence of the model structure and then the judgment method of critical instability of the model structure is given. Secondly, the improved method of the control system is given when there is critical instability of the model structure caused by identification error and mismatch between process control system design and equipment. Simulation results validate the influence of the model structure on the control performance and the effectiveness of the improved method.