机电工程技术2025，Vol.54Issue(14)：40-45,6.DOI:10.3969/j.issn.1009-9492.2024.00117
基于贝叶斯优化的脱硝过程RTO-MPC自主优化研究
Self-optimization of RTO-MPC for Denitrification Process Based on Bayesian Optimization
摘要
Abstract
Real-time optimization(RTO)and model predictive control(MPC)are extensively applied in industrial processes to achieve optimal economic objectives.However,a significant challenge arises when the optimization objectives of RTO and the models related to process dynamics are both unknown.To address this issue,a self-optimization method based on Bayesian optimization(BO)is proposed.When the RTO objective and process dynamics are measurable but the mechanisms or explicit models are unknown,the RTO objective is treated as a black-box one,and the steady-state condition of the process is considered black-box equality constraints.First,Gaussian process(GP)models are employed to approximate the black-box objective and constraints.Secondly,a constrained BO method is developed to optimize the setpoints iteratively by treating various operating conditions as contextual information,accounting for the effects of measurable disturbances.For the MPC layer,the GP model corresponding to the process steady-state conditions is used as the predictive model,and another contextual mechanism is utilized to automatically optimize the MPC controller parameters,enhancing the tracking performance of the MPC controller.Simulation results on the denitrification process demonstrates that the data-driven approach can optimize both setpoints and the MPC parameters in real-time,adapting to fluctuating conditions without prior knowledge of system dynamics and optimization objective.This offers a novel perspective for the self-optimization of industrial control systems in complex environments.关键词
实时优化/模型预测控制/贝叶斯优化/脱硝过程Key words
real-time optimization/model predictive control/Bayesian optimization/denitrification process分类
资源环境引用本文复制引用
龚琨,张凌智,曹振斌,李鹏飞,田雁斌,张琳,谢磊..基于贝叶斯优化的脱硝过程RTO-MPC自主优化研究[J].机电工程技术,2025,54(14):40-45,6.基金项目
国家自然科学基金(62073286) （62073286）
