文献类型：期刊文献

中文题名：基于深度学习的鲁棒非线性模型预测控制方法

英文题名：Robust nonlinear model predictive control method based on deep learning

作者：孙京诰[1];陈显锋[1];李郅辰[1]

机构：[1]华东理工大学信息科学与工程学院,上海200237

年份：2022

卷号：43

期号：6

起止页码：1762

中文期刊名：计算机工程与设计

外文期刊名：Computer Engineering and Design

收录：CSTPCD;;北大核心:【北大核心2020】；

基金：国家自然科学青年基金项目(61803159)。

语种：中文

中文关键词：非线性模型预测控制;深度学习;鲁棒控制器;神经网络;半间歇聚合反应器

外文关键词：nonlinear model predictive control;deep learning;robust control;neural network;semi-batch polymerization reactor

摘要：为降低鲁棒非线性模型预测控制方法中优化问题的实时求解难度,提出一种基于深度学习的近似鲁棒控制器方法。利用复杂的鲁棒非线性模型预测控制算法作为训练数据的生成器,以当前时刻的过程状态作为网络的输入,复杂控制算法计算的最优控制输入作为网络的输出,基于深度神经网络学习复杂的非线性模型预测控制策略。通过一个工业半间歇聚合反应器模型案例验证了所提方法的有效性,深层网络与浅层网络相比具有更好的效果。
To reduce the difficulty of real-time solving optimization problem in the robust nonlinear model predictive control method,an approximate robust controller method based on deep learning was proposed.A complex robust nonlinear model predictive control algorithm was used as a generator of training data,the current process state was used as network input,the optimal control input calculated using the complex control algorithm was used as network output,and complex nonlinear model predictive control strategies were learnt based on deep neural networks.An industrial semi-batch polymerization reactor model case verifies the effectiveness of the proposed method,and the deep network has better results than the shallow network.