基于小波网络模型的EPI过程的控制和优化

Control and Optimization of EPI Reaction Ristillation Process Based on Wavelet Networks

针对具有强非线性、大纯滞后特性的EPI反应分馏生产过程的废液DOC体积分数控制这一控制难题,应用小波神经网络模型和结合机理分析模型及小波神经网络模型的混合模型结构进行实际EPI反应分馏生产过程的建模控制和优化,提出了基于稳态小波神经网络模型的优化和基于动态小波神经网络模型的非线性预测控制两层结构的整体解决方案,在实际生产过程中获得了成功的应用,不仅使生产过程平稳运行,而且显著地降低排放废液中的DOC体积分数,达到了环保的标准。所提出的技术适合于许多连续反应系统。

Aiming at the difficult problem about controlling the concentration of dissolved organic carbon(DOC) in the bottom outlet of the EPI reaction distillation tower due to strong nonlinearity and large time dead-delay, the wavelet network model and hybrid network structure incorporating first-principle knowledge and wavelet network are applied to the modeling,controlling and optimizing of EPI reaction distillation process.The total solution with two layers structure strategy making up of optimization based on steady wavelet network and nonlinear predictive control based on dynamic wavelet network are proposed and applied to a real production process successfully.Not only the production process operates more steadily,but also DOC concentration is reduced obviously and the allowing standard is reached.The proposed technique is applicable to many continuous reaction systems.