高等学校化学学报 ›› 2024, Vol. 45 ›› Issue (10): 20240296.doi: 10.7503/cjcu20240296

• 物理化学 • 上一篇    下一篇

基于机器学习的CH4燃烧动力学机理优化

曹双双1, 黄济勇1, 李伟2, 张厚君1, 李象远3, 韩优1()   

  1. 1.天津大学化工学院, 天津 300072
    2.航天化学能源全国重点实验室, 湖北航天化学技术研究所, 襄阳 441003
    3.四川大学化学工程学院, 成都 610065
  • 收稿日期:2024-06-19 出版日期:2024-10-10 发布日期:2024-08-12
  • 通讯作者: 韩优 E-mail:yhan@tju.edu.cn
  • 作者简介:第一联系人:共同第一作者.
  • 基金资助:
    航天化学能源全国重点实验室开放基金(STACPL220221B03);国家自然科学基金(T2441001)

Optimization of Kinetic Mechanism for Methane Combustion Based on Machine Learning

CAO Shuangshuang1, HUANG Jiyong1, LI Wei2, ZHANG Houjun1, LI Xiangyuan3, HAN You1()   

  1. 1.School of Chemical Engineering and Technology,Tianjin University,Tianjin 300072,China
    2.National Key Laboratory of Aerospace Chemical Power,Hubei Institute of Aerospace Chemotechnology,Xiangyang 441003,China
    3.School of Chemical Engineering,Sichuan University,Chengdu 610065,China
  • Received:2024-06-19 Online:2024-10-10 Published:2024-08-12
  • Contact: HAN You E-mail:yhan@tju.edu.cn
  • Supported by:
    the Open Research Found Program of National Key Laboratory of Aerospace Chemical Power, China(STACPL220221B03);the National Natural Science Foundation of China(T2441001)

摘要:

基于径向基函数插值算法构建的机器学习模型, 以点火延迟时间(T=1084~2175 K, p=7.3×104~2.4×106 Pa, φ=0.2~2.0)和层流火焰速度(T=293~600 K, p=5.1×104~1.1×106 Pa, φ=0.4~2.0)实验数据为优化目标, 对CH4燃烧机理的指前因子和活化能进行优化, 获得了可在宽工况范围内使用的CH4燃烧机理. 与初始的CH4机理 (Ori-CH4)相比, 优化后的CH4机理(Opt-CH4)在点火延迟时间上的预测平均误差下降了57.46%, 在层流火焰速度上的预测平均误差下降了21.55%. 使用Opt-CH4机理对点火延迟时间、 层流火焰速度和射流搅拌反应器中的组分浓度变化趋势进行了预测, Opt-CH4机理均表现出优越的预测准确度. 在T=1491.5 K, p=1.0×105 Pa, 4.988%CH4\19.953%O2\75.059%N2(体积分数)工况下, CH3+O2CH2O+OH和CH2O+O2HCO+HO2在各个机理中的敏感性差异是优化前后CH4机理预测准确度不同的主要原因. 因此, 机器学习方法在燃料燃烧反应动力学机理参数优化上具有广阔的应用前景.

关键词: 甲烷燃烧, 机器学习, 化学动力学, 机理优化

Abstract:

In this work, the experimental data of ignition delay time(T=1084—2175 K, p=7.3×104—2.4×106 Pa, φ=0.2—2.0) and laminar flame speed(T=293—600 K, p=5.1×104—1.1×106 Pa, φ=0.4—2.0) were taken as the optimization objectives based on the machine-learning model constructed by radial basis function interpolation method, and pre-exponential factors and activation energies of CH4 combustion mechanism were optimized, and a CH4 combustion mechanism that can be used in a wide range of working conditions was obtained. Compared with the Ori-CH4 mechanism, the mean error of the Opt-CH4 mechanism is reduced by 57.46% in the ignition delay times and 21.55% in the laminar flame speeds. The Opt-CH4 mechanism was used to predict the ignition delay times, laminar flame speeds and the variation tendency of species concentration in jet stirred reactor. The Opt-CH4 mechanism showed superior prediction accuracy. Under the conditions of T=1491.5 K, p=1.0×105 Pa, 4.988%CH4\19.953%O2\75.059%N2(volume fraction), the difference of sensitivity of CH3+O2CH2O+OH and CH2O+O2HCO+HO2 in each mechanism is the main reason for the difference of prediction accuracy of CH4 mechanism before and after optimization. Therefore, the machine learning method has a broad application prospect in the optimization of fuel combustion reaction kinetics mechanism parameters.

Key words: Methane combustion, Machine learning, Chemical kinetics, Mechanism optimization

中图分类号: 

TrendMD: