带约束多项式权重LPV模型辨识算法

Identification Method for LPV Models Using Polynomial Weights with Constraints

针对复杂工业非线性系统建模难度大、精度低等问题,基于多模型插值的变参数线性(LPV)模型辨识算法,提出双工作点变量条件下的带约束多项式权重函数结构及其参数寻优策略,以有效降低非线性系统辨识的难度并充分保证所建模型的全局稳定性.以高纯度分馏塔这一典型非线性工业过程为研究对象进行LPV模型建模与仿真,获得了较好的输出和阶跃响应曲线拟合结果,验证了LPV模型能够充分反映非线性系统的运行特性以及所提算法的有效性和实用性.

In order to better handle the complex nonlinear industrial processes and improve model accuracy, this paper studied the identification algorithm for LPV （ linear parameter varying） models based on multi-model interpolation. The structure and the corresponding optimization strategy of two-scheduling-variables-based polynomial weighting function with constrains were proposed, which can effectively reduce the difficulty of nonlinear model identification and guarantee the model’s global stability. The accuracy and effectiveness of the developed LPV model were verified by simulating a typical industrial process, namely a high purity distillation column. Both the LPV model outputs and step responses fit the real process well.