基于数据的间歇过程时变神经模糊模型研究

Research on data-based time-varying neuro-fuzzy model for batch processes

间歇过程的优化控制往往依赖于过程精确的数学模型,快速反应的市场要求使得数据驱动的建模方法被应用到了间歇过程的建模中。但常规的数据驱动建模方法在模型结构中没有考虑间歇过程具有重复性的特性,只是简单地将间歇过程作为一般的非线性结构进行处理。针对该问题,本文提出一种新颖的间歇过程时变神经模糊模型,将时间轴和批次轴的信息统一在二维集成模型的框架下,对间歇过程的输入输出数据按照三维信息进行处理,使模型参数变为时间的函数,从而按照批次轴方向进行学习,合理地应用了间歇过程在批次轴方向上的重复性信息。因此,通过引入时变权重的概念,使模型结构中蕴含了间歇过程重复性的特性。在此基础上,提出一种基于迭代学习和Lyapunov方法的参数学习算法,避免了传统学习算法中学习参数采用试凑法的缺点,并对模型的收敛性进行了严格的数学分析,得出模型的学习参数在批次轴方向上渐进收敛的结论。最后,将本文提出的时变神经模糊模型应用到一典型间歇过程的建模研究中,与传统的神经模糊模型进行了对比,仿真研究表明,本文提出的时变神经模糊模型具有较好的非线性逼近和自学习能力,能够反应间歇过程的二维模型特性,为间歇过程的建模研究提供了一条新的途径。

The optimization control of batch process usually depends on accurate mathematical model and the data-driven method has to be a powerful modeling tool for batch process due to the rapid market requirement. But the characteristic of repetitive was not considered in the structure of traditional data-deriven model, which just simply looks the batch process as a general nonlinear process. Thus, a novel time-varying neuro-fuzzy model is proposed in this paper, which combines the information of time and batch into one integrated model structure. The input and output datas of batch process are looked as three dimension information. Based on it, the model parameters are the functions of time and then are updated with the direction of batch by using the repetitive information. Thus it uses the concept of time-varying weights which implies the characteristic of repetitive. Based on this structure, a new algorithm based on Lyapunov method and iterative learning algorithm is presented, which avoids the drawbacks of tried-and-error. Moreover, the rigorous proof is given to verify the convergence of time-varying neuro-fuzzy model and the conclusion that the changes of learning parameters converge with respect to the batch index number is gotten. The simulation results show that the time-varying neuro-fuzzy model has strong ability to approximate any nonlinear system with two-dimension characteristic, and nrovide a new way for modeling, of batch orocess.