自动化学报2025，Vol.51Issue(10)：2163-2177,15.DOI:10.16383/j.aas.c250272
基于强化学习的流程工业智能决策研究与展望
A Review and Perspective on Reinforcement Learning for Intelligent Decision-making in Process Industries
摘要
Abstract
Process industries constitute a vital component of modern manufacturing systems,where decision-mak-ing for process optimization directly impacts both economic performance and resource utilization efficiency.As pro-duction scale expands and system complexity increases,traditional optimization approaches relying on mechanism-based modeling or heuristic rules are gradually revealing their limitations in handling the high-dimensional coupling,nonlinearity,and uncertainty inherent to industrial processes.Reinforcement learning,with its model-free nature and capabilities in efficient decision-making,adaptive adjustment,and uncertainty handling,holds promise for ad-dressing these challenges and has emerged as a key research direction for intelligent decision-making in process in-dustries.However,the practical implementation of reinforcement learning in process industries still faces challenges,including the high dimensionality and structural diversity of state-action spaces,complex process constraints,and strong non-stationarity of operating conditions.This paper systematically reviews current applications and key tech-nologies of reinforcement learning in process industries.It focuses on algorithmic advances and application develop-ments in complex decision-making spaces,constraint handling,large-scale systems,and uncertain environments.Fi-nally,it outlines future development trends and potential research directions,aiming to provide a theoretical found-ation and methodological support for the intelligent optimization of complex industrial systems.关键词
强化学习/流程工业/大规模/不确定性Key words
Reinforcement learning/process industries/large-scale/uncertainty引用本文复制引用
黄慕轶,朱佳雯,戴鑫,杜文莉,钱锋..基于强化学习的流程工业智能决策研究与展望[J].自动化学报,2025,51(10):2163-2177,15.基金项目
国家重点研发计划(2022YFB3305900),国家自然科学基金(62136003,62394343,62394345),中央高校基本科研业务费专项资金(222202517006)资助Supported by National Key Research and Development Pro-gram of China(2022YFB3305900),National Natural Science Foundation of China(62136003,62394343,62394345),and Fun-damental Research Funds for the Central Universities(222202517006) （2022YFB3305900）
