基于深度强化学习的真空回潮机多相流传质传热过程智能调控方法

doi:10.13385/j.cnki.vacuum.2026.05.12

摘要/Abstract

摘要： 在动态变化的低压环境中,真空回潮工艺可以实现对多相流过程的精确调控,温湿度仅是调控过程中的宏观表征。传统真空回潮控制因真空环境的时变性、非线性和强耦合性,在粗抽、精抽、保压加潮等不同工艺阶段难以建立精确机理模型,导致控制精度不足。为此提出基于深度强化学习的真空回潮机多相流传质传热过程智能调控方法。设定真空回潮机温湿度多目标控制约束条件,结合温湿度耦合关系模拟其动态演化过程;基于DQN智能体强大的环境感知和决策能力,自适应地学习并逼近真空环境下多相流传质传热的复杂动力学模型,生成最优控制策略;调节蒸汽阀门开度和蒸汽压力,实现真空回潮机温湿度多目标优化控制。实验结果表明,与传统控制方法（取两者结果均值）相比,优化设计方法使温度超调量和湿度超调量分别平均降低0.83个和1.8个百分点,温湿度震荡幅值分别平均降低0.36℃和0.42个百分点,平均多目标冲突指数从0.28降至0.03。这证明优化设计方法不仅在控制精度和协调度方面具有显著优势,还为复杂真空工艺过程的智能化、精细化调控提供了一种数据驱动的新范式。

关键词: 真空回潮, 真空工艺, 多相流传质, 蒸汽喷射系统

Abstract: In a dynamically changing low-pressure environment, vacuum conditioning technology can achieve precise control of multiphase flow processes, and temperature and humidity are only macroscopic representations of the control process. Due to the time-varying, nonlinear, and strongly coupled nature of the vacuum environment, it is difficult for traditional vacuum conditioning control to establish accurate mechanism models at different stages of the process, such as rough pumping, fine pumping, and pressure and humidity control, resulting in insufficient control accuracy. A deep reinforcement learning based intelligent control method for multiphase mass and heat transfer processes in vacuum conditioning is proposed for this purpose. It sets multi-objective control constraints for temperature and humidity of the vacuum conditioning, and simulates the dynamic evolution process by combining the coupling relationship between temperature and humidity. Based on the powerful environmental perception and decision-making capabilities of DQN intelligent agents, it adaptively learns and approximates complex dynamic models of multiphase flow mass and heat transfer in vacuum environments, and generates optimal control strategies. It then adjusts the opening of the steam valve and steam pressure to achieve multi-objective optimization control of temperature and humidity in the vacuum conditioning. The experimental results demonstrate that, compared with traditional control methods(mean of two test results), the optimized design approach reduces the temperature and humidity overshoots by an average of 0.83 and 1.8 percentage points, respectively; decreases the average oscillation amplitude of temperature and humidity by 0.36℃ and 0.42 percentage points, respectively; and lowers the average multi-objective conflict index from 0.28 to 0.03. This proves that the optimized design method not only has significant advantages in control accuracy and coordination, but also provides a data-driven new paradigm for intelligent and refined control of complex vacuum processes.

Key words: vacuum conditioning, vacuum process, multiphase mass transfer, steam injection system

中图分类号: TP399

高铁功, 杨兴权. 基于深度强化学习的真空回潮机多相流传质传热过程智能调控方法[J]. 真空, 2026, 63(5): 81-88.

GAO Tiegong, YANG Xingquan. Intelligent Control Method for Multiphase Mass and Heat Transfer Process of Vacuum Conditioning Machine Based on Deep Reinforcement Learning[J]. VACUUM, 2026, 63(5): 81-88.

参考文献

[1] 邓雷,崔光磊.基于COMSOL Multiphysics软件的某地铁配电柜温湿度控制策略[J].城市轨道交通研究,2025,28(4):71-77.
[2] 陈长成,安晶晶,王闯,等.基于强化学习的室内温湿度联合控制方法研究[J].科学技术与工程,2024,24(12):5123-5133.
[3] 赵海波,张静峰,吴坤,等.基于预测PI的热泵干燥温湿度解耦控制研究[J].控制工程,2024,31(12):2149-2158.
[4] 岑江晖,何德峰,朱威.基于数据建模和参数压缩的连续干燥过程温湿度预测控制[J].高技术通讯,2023,33(7):713-721.
[5] 郝玉山,李新叶,杨海涛,等. 烟叶真空回潮的建模与控制[J].中国烟草学报,2019,25(4):29-35.
[6] 裴志明,徐娟,范晓宝,等.复合式真空回潮机的设计及试验[J].烟草科技,2016,49(9):87-91.
[7] 王清,李琮琮,荆臻,等.基于滑动迭代离散傅里叶的并网设备多目标协同控制方法研究[J].电测与仪表,2023,60(3):157-164.
[8] Jiang Tao, Yu Fei, Jiang Qianmin, et al.Study on the mass transfer characteristics of hydrogen in heavy oil by a modified dynamic pressure step method[J].Petroleum Science, 2025,22(3):1360-1369.
[9] Zhu Min, Ping Yuchen, Zhang Yinghao, et al.Discharge and mass transfer characteristics of atmospheric pressure gas-solid two-phase gliding arc[J].Plasma Science and Technology,2024,26(9):88-96.
[10] 刘卓瑶,常建华,倪海彬,等.基于表面等离子体共振效应的单模光纤高灵敏度温湿度传感器[J].中国激光,2023,50(14):193-201.
[11] Wei Xiaoxinao, Zhang Renhui.The axial tip clearance leakage analysis of the winglet and composite blade tip for the liquid-ring vacuum pump[J]. Vacuum, 2022, 200:111027.
[12] Zhang Junzheng, Xu Zhen.Design research of a novel aftercool-humidifier concept for humid air turbine cycle[J].Journal of Thermal Science,2024,33(3):951-969.
[13] 王志梁,严立辰,夏宇栋,等.直膨空调系统动态建模及温湿度解耦控制[J].低温与超导,2023,51(7):63-70,81.
[14] Li Xu, Zhang Dongwei, Chen Si, et al.Outstanding proton conductivity over wide temperature and humidity ranges and enhanced mechanical,thermal stabilities for surface-modified MIL-101-Cr-NH₂/Nafion composite membranes[J].Green Energy & Environment,2024,9(11):1734-1746.
[15] 韩晓婉,邹同华,陈剑波,等.LiCl溶液真空蒸发过程的实验研究[J].真空科学与技术学报, 2019, 39(9):775-780.
[16] 郭宁,李晓青.基于自适应模糊PID的电气设备温度调试仿真[J].计算机仿真, 2024, 41(2):56-60,119.
[17] Li Shuxun, Wang Yiting, Liu Dan, et al.Noise-reducing structure optimization of inverted bucket steam valves based on SVM-NOA[J]. Acoustics,2025,7(4):74.
[18] 李晓敏,王立军.散热器热量传递特性的研究[J].真空科学与技术学报, 2018, 38(1):71-74.
[19] 张治华,李晓林,杨嘉,等.抽真空过程中水面蒸发及空气温湿度变化特性研究[J].真空,2025,62(5):1-10.
[20] 汤嘉伟,王成轩,吴征天.基于SMC的变风量空调房间温湿度解耦控制[J].控制工程,2023,30(1):162-168.
[21] Zhao Quanbin, Fu Pengfei, Chong Daotong, et al.Steam entrainment and heat transfer characteristic of steam submerged jet condensation within steam-water mixture layer[J].International Journal of Heat and Mass Transfer,2023,201:123600.
[22] 潘朱良,刘昇,毛海镰,等.真空回潮工序连续生产的改进设计与应用[J].设备管理与维修,2024(7):60-63.
[23] 程移风,黄重辉,章毅,等.真空回潮机箱体内输送链张紧装置的分析[J].中国机械,2023(17):44-47.
[24] 董晓斌,姚明笙,刘文,等.真空回潮机气动球阀无法开启故障分析与修复[J].农业装备技术,2023,49(2):25-27.
[25] Dai Lingrui.Multi-objective optimization design of vacuum interrupter structure based on improved artificial jellyfish search algorithm[J]. IEEJ Transactions on Electrical and Electronic Engineering,2023,18(11):1806-1814.
[26] Bohacek J, Karimi-Sibaki E, Vakhrushev A, et al.A coupled magnetohydrodynamics (MHD) and thermal stress-strain model to explore the impact of gas cooling on ingot solidification shrinkage in vacuum arc remelting (VAR) process[J]. Metallurgical and Materials Transactions, B. Process metallurgy and materials processing science,2024, 55(6):4408-4417.
[27] Kong Xiaoyu, Cao Yun, Zhu Hengbo, et al.Experimental and numerical study of MEMS electrothermal actuators: comparing dynamic behavior and heat transfer process in vacuum and non-vacuum environments[J].Vacuum, 2024, 227:113409.

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed