煤炭超临界水制氢反应器内多相流场智能滚动预测研究

doi:10.11949/0438-1157.20240228

摘要/Abstract

摘要：

煤炭超临界水制氢技术在高温高压条件下利用超临界水充分气化煤炭，实现了高效低排放的转化和制氢过程。为解决因反应器内复杂多相流行为导致的仿真耗时问题，以及常见代理模型时序预测时间短、精度下降快等问题，提出基于本征正交分解（proper orthogonal decomposition，POD）和Koopman理论的深度学习模型POD-Koopman，用于捕捉和学习反应器内复杂流场的长时时空演变特征，实现数据驱动的长时滚动预测。测试结果表明其能在较小计算开销下准确滚动预测反应器内多相流场时变行为，助力下游制氢反应器工业化设计及优化任务。

关键词: 超临界水煤制氢, 反应器, 本征正交分解, Koopman, 瞬态多相流, 长时滚动预测

Abstract:

The coal-supercritical water hydrogen production technology utilizes high-temperature and high-pressure conditions to gasify coal in supercritical water, enabling efficient and low-emission coal conversion and hydrogen production process. To alleviate the time-consuming simulation process caused by the complex multiphase flow behavior within the reactor, as well as the issues such as short prediction time and rapid prediction deterioration when constructing conventional data-driven surrogate models, this paper proposed a deep learning model called POD-Koopman based on the proper orthogonal decomposition (POD) and the Koopman theory, which can capture and learn the long-term spatiotemporal evolution characteristics of the complex flow, thus facilitating the long-term rolling prediction. The test results show that it can accurately predict the time-varying behavior of the multiphase flow field in the reactor on a rolling basis with a small computational overhead, and assist in the industrial design and optimization tasks of downstream hydrogen production reactors.

Key words: coal-supercritical water hydrogen production, reactor, proper orthogonal decomposition, Koopman, transient multiphase flow, long-term rolling prediction

中图分类号:

TK 123

丁家琦, 刘海涛, 赵普, 朱香凝, 王晓放, 谢蓉. 煤炭超临界水制氢反应器内多相流场智能滚动预测研究[J]. 化工学报, 2024, 75(8): 2886-2896.

Jiaqi DING, Haitao LIU, Pu ZHAO, Xiangning ZHU, Xiaofang WANG, Rong XIE. Study on intelligent rolling prediction of the multiphase flows in coal-supercritical water fluidized bed reactor for hydrogen production[J]. CIESC Journal, 2024, 75(8): 2886-2896.

图/表 12

图1 Koopman动力学演变示意图

Fig.1 The illustration of state evolution through the Koopman theory

图2 用于煤炭超临界水制氢反应器多相流场长时滚动预测的POD-Koopman框架

Fig.2 The POD-Koopman framework for long-term rolling prediction of multiphase flow fields in the reactor

图3 反应器几何模型、二维结构化网格以及气化产物摩尔分数场

Fig.3 The reactor geometry, 2D structured mesh grids, and mole fraction fields of gasification products in the reactor

表1 反应器仿真产气量与实验值对比

Table 1 Comparison of simulated gas production of the reactor with experimental values

项目	H₂	CO₂	CO	CH₄
仿真值/(kg/mol)	37.82	20.86	2.67	2.22
实验值/(kg/mol)	39.83	21.52	3.21	0.84

表2 POD-Koopman模型训练配置

Table 2 The POD-Koopman model training configuration

区块	子模型	超参数	张量维度
输入	—	—	[500, 512, 32]
输入	Reshape	—	[500, 16384]
降维	POD	Rank=200	[500, 200]
编码	MLP (ReLU)	[200, 128]	[500, 128]
	MLP (ReLU)	[128, 128]	[500, 128]
	MLP	[128, 64]	[500, 64]
Koopman递推	MLP (b = False)	[64, 64]	[500, 64]
解码	MLP (ReLU)	[64, 128]	[500, 128]
	MLP (ReLU)	[128, 128]	[500, 128]
	MLP	[128, 200]	[500, 200]
还原	POD		[500, 16384]
输出	Reshape	—	[500, 512, 32]

图4 反应器气化产物POD分解结果

Fig.4 The POD processing result of gasification products in the reactor

图5 POD-Koopman长时滚动预测结果

Fig.5 The long-term rolling prediction result of the POD-Koopman model

图6 POD-Koopman长时滚动预测结果

Fig.6 The long-term rolling prediction results of the POD-Koopman model

图7 不同训练集长度下POD-Koopman模型滚动预测精度变化

Fig.7 The variation of rolling-out prediction accuracy of POD-Koopman with different training set lengths

表3 3种预测模型基本配置比较

Table 3 The configuration comparison of three predictive models

模型处理器	POD-Koopman	MLP	LSTM
损失函数	潜空间线性误差+重构误差	重构误差	重构误差
递推格式	RK-4	—/—	—/—
训练耗时/s	593	272	437

图8 不同模型滚动预测精度衰减至PSNR=30 dB阈值时所预测的H2摩尔分数场及时刻

Fig.8 The comparison of predicted H2 mole fractional field as well as the time stamp of different models when the rolling prediction PSNR is 30 dB

表4 不同模型在外推测试集（48~70 s）上滚动预测的PSNR指标的均值和标准差

Table 4 Mean and standard deviation of the PSNR metric of predictions of different models on the test set （48—70 s）

PSNR	POD-Koopman	MLP	LSTM
H₂	$31.9519 ± 6.6336$	$26.4810 ± 1.8832$	$30.8269 ± 3.5498$
CO₂	$29.3640 ± 3.7685$	$29.3177 ± 1.4964$	$27.9130 ± 3.9616$
CO	$28.2562 ± 4.1335$	$28.3082 ± 0.6031$	$30.4060 ± 3.1454$
CH₄	$35.8227 ± 5.9926$	$33.5829 ± 4.0012$	$34.6492 ± 5.0899$
Vol	$37.3822 ± 2.6248$	$34.2748 ± 2.8371$	$36.2867 ± 4.4020$

表4 不同模型在外推测试集（48~70 s）上滚动预测的PSNR指标的均值和标准差

Table 4 Mean and standard deviation of the PSNR metric of predictions of different models on the test set （48—70 s）

PSNR	POD-Koopman	MLP	LSTM
H₂	$31.9519 ± 6.6336$	$26.4810 ± 1.8832$	$30.8269 ± 3.5498$
CO₂	$29.3640 ± 3.7685$	$29.3177 ± 1.4964$	$27.9130 ± 3.9616$
CO	$28.2562 ± 4.1335$	$28.3082 ± 0.6031$	$30.4060 ± 3.1454$
CH₄	$35.8227 ± 5.9926$	$33.5829 ± 4.0012$	$34.6492 ± 5.0899$
Vol	$37.3822 ± 2.6248$	$34.2748 ± 2.8371$	$36.2867 ± 4.4020$

参考文献 32

8	寇家庆, 张伟伟. 动力学模态分解及其在流体力学中的应用[J]. 空气动力学学报, 2018, 36(2): 163-179.
	Kou J Q, Zhang W W. Dynamic mode decomposition and its applications in fluid dynamics[J]. Acta Aerodynamica Sinica, 2018, 36(2): 163-179.
9	张伟伟, 寇家庆, 刘溢浪. 智能赋能流体力学展望[J]. 航空学报, 2021, 42(4): 524689.
	Zhang W W, Kou J Q, Liu Y L. Prospect of artificial intelligence empowered fluid mechanics[J]. Acta Aeronautica et Astronautica Sinica, 2021, 42(4): 524689.
10	王烨, 朱欣悦, 孙振东. 基于POD降阶模型的正弦波翅片扁管管翅式换热器流动与传热特性分析[J]. 化工学报, 2022, 73(5): 1986-1994.
	Wang Y, Zhu X Y, Sun Z D. Flow and heat transfer characteristics analysis of flat tube-bank-fin heat exchanger with sine wave fin based on POD reduced-order model[J]. CIESC Journal, 2022, 73(5): 1986-1994.
11	籍帅航, 王金江, 蔡睿, 等. 数字孪生驱动的热交换器降阶建模及智能感知方法研究[J]. 化工学报, 2023, 74(10): 4218-4228.
	Ji S H, Wang J J, Cai R, et al. Research on reduced order modeling and intelligent sensing method for heat exchangers driven by digital twin[J]. CIESC Journal, 2023, 74(10): 4218-4228.
12	可钊, 时永鑫, 张鹏, 等. 基于全局POD降阶模型的复杂薄壁结构减振优化[J]. 航空学报, 2023, 44(13): 227900.
	Ke Z, Shi Y X, Zhang P, et al. Vibration reduction optimization of complex thin-walled structures based on global POD reduced-order model[J]. Acta Aeronautica et Astronautica Sinica, 2023, 44(13): 227900.
13	Liu S K, Yang Y, Yu L J, et al. Predicting gas production by supercritical water gasification of coal using machine learning[J]. Fuel, 2022, 329: 125478.
14	Ma Z R, Wang J J, Feng Y S, et al. Hydrogen yield prediction for supercritical water gasification based on generative adversarial network data augmentation[J]. Applied Energy, 2023, 336: 120814.
15	Xie X Y, Wang X F, Zhao P, et al. Learning time-aware multi-phase flow fields in coal-supercritical water fluidized bed reactor with deep learning[J]. Energy, 2023, 263: 125907.
16	Hao Y C, Xie X Y, Zhao P, et al. Forecasting three-dimensional unsteady multi-phase flow fields in the coal-supercritical water fluidized bed reactor via graph neural networks[J]. Energy, 2023, 282: 128880.
17	Koopman B O. Hamiltonian systems and transformation in Hilbert space[J]. Proceedings of the National Academy of Sciences of the United States of America, 1931, 17(5): 315-318.
18	Bevanda P, Sosnowski S, Hirche S. Koopman operator dynamical models: learning, analysis and control[J]. Annual Reviews in Control, 2021, 52: 197-212.
19	Otto Samuel E, Rowley Clarence W. Koopman operators for estimation and control of dynamical systems[J]. Annual Review of Control, Robotics, and Autonomous Systems, 2021, 4: 59-87.
20	Brunton S L, Budišić M, Kaiser E, et al. Modern Koopman theory for dynamical systems[J]. SIAM Review, 2022, 64(2): 229-340.
21	Takeishi N, Kawahara Y, Yairi T. Learning Koopman invariant subspaces for dynamic mode decomposition[C]//Proceedings of the 31st International Conference on Neural Information Processing Systems. Long Beach, California, USA, ACM, 2017: 1130-1140.
22	Navaneeth N, Chakraborty S. Koopman operator for time-dependent reliability analysis[J]. Probabilistic Engineering Mechanics, 2022, 70: 103372.
23	Chen Z Y, Lin Z W, Zhai X Y, et al. Dynamic wind turbine wake reconstruction: a Koopman-linear flow estimator[J]. Energy, 2022, 238: 121723.
24	Wang R, Dong Y H, Arik S Ö, et al. Koopman neural forecaster for time series with temporal distribution shifts[J/OL]. ArXiv, .
25	Schmid P J. Dynamic mode decomposition of numerical and experimental data[J]. Journal of Fluid Mechanics, 2010, 656: 5-28.
26	Rumelhart D E, Hinton G E, Williams R J. Learning representations by back-propagating errors[J]. Nature, 1986, 323: 533-536.
1	Guo L J, Jin H. Boiling coal in water: hydrogen production and power generation system with zero net CO₂ emission based on coal and supercritical water gasification[J]. International Journal of Hydrogen Energy, 2013, 38(29): 12953-12967.
2	Jin H, Liu S K, Wei W W, et al. Experimental investigation on hydrogen production by anthracene gasification in supercritical water[J]. Energy & Fuels, 2015, 29(10): 6342-6346.
3	Sun J L, Feng H F, Kou J J, et al. Experimental investigation on carbon microstructure for coal gasification in supercritical water[J]. Fuel, 2021, 306: 121675.
4	郭斯茂, 郭烈锦, 聂立, 等. 超临界水流化床内煤气化过程建模与仿真(1): 数学模型及物理场分布规律[J]. 工程热物理学报, 2014, 35(3): 507-511.
	Guo S M, Guo L J, Nie L, et al. Modeling and numerical study on a supercritical water fluidized bed reactor for coal gasification(1): Mathematical model and distributions of physical fields[J]. Journal of Engineering Thermophysics, 2014, 35(3): 507-511.
5	郭斯茂, 郭烈锦, 聂立, 等. 超临界水流化床内煤气化过程建模与仿真(2): 气化反应动力学模型及气化规律[J]. 工程热物理学报, 2014, 35(12): 2429-2432.
	Guo S M, Guo L J, Nie L, et al. Modeling and numerical study on a supercritical water fluidized bed reactor for coal gasification(2): Gasification kinetic model and gasification characters[J]. Journal of Engineering Thermophysics, 2014, 35(12): 2429-2432.
6	Ren Z H, Guo L J, Jin H, et al. Integration of CFD codes and radiation model for supercritical water gasification of coal in fluidized bed reactor[C]//International heat transfer conference digital library. Begel House Inc., 2018.
7	任振华, 金辉, 刘石, 等. 煤炭超临界水流化床制氢反应器内颗粒流动及传热特性的数值分析[J]. 工程热物理学报, 2020, 41(1): 154-160.
	Ren Z H, Jin H, Liu S, et al. Numerical analysis of particle flow and heat transfer characteristics in a coal-supercritical water fluidized bed reactor for hydrogen production[J]. Journal of Engineering Thermophysics, 2020, 41(1): 154-160.
27	Lumley J L. The structure of inhomogeneous turbulent flows[J]. Atmospheric Turbulence and Radio Wave Propagation, 1967: 166-178.
28	Su X H, Guo L J, Jin H. Mathematical modeling for coal gasification kinetics in supercritical water[J]. Energy & Fuels, 2016, 30(11): 9028-9035.
29	Wang Z, Bovik A C, Sheikh H R, et al. Image quality assessment: from error visibility to structural similarity[C]// IEEE Transactions on Image Processing. IEEE, 2004: 600-612.
30	Zeng A L, Chen M X, Zhang L, et al. Are transformers effective for time series forecasting?[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2023, 37(9): 11121-11128.
31	Vaswani A, Shazeer N, Parmar N, et al. Attention is all you need[C]// Proceedings of the 31st International Conference on Neural Information Processing Systems. Long Beach, California, USA, ACM, 2017: 6000-6010.
32	Hochreiter S, Schmidhuber J. Long short-term memory[J]. Neural Computation, 1997, 9(8): 1735-1780.

[1]	王芝安, 兰忠, 马学虎. 喷嘴参数对超临界水热燃烧特性影响的模拟[J]. 化工学报, 2024, 75(6): 2190-2200.
[2]	王文雅, 张玮, 楼小玲, 钟若菲, 陈冰冰, 贠军贤. 纳米纤维素嵌合型晶胶微球的多微管成形与模拟[J]. 化工学报, 2024, 75(5): 2060-2071.
[3]	薛潇, 商敏静, 苏远海. 微反应器内药物连续流合成的研究进展[J]. 化工学报, 2024, 75(4): 1439-1454.
[4]	韩宇, 周乐, 张鑫, 罗勇, 孙宝昌, 邹海魁, 陈建峰. 高黏附性Pd/SiO₂/NF整体式催化剂的制备及加氢性能研究[J]. 化工学报, 2024, 75(4): 1533-1542.
[5]	何宇航, 谢丹, 吕阳成. 微反应器内阳离子聚合研究进展[J]. 化工学报, 2024, 75(4): 1302-1316.
[6]	王成秀, 宋大山, 李之辉, 杨潇, 蓝兴英, 高金森, 徐春明. Geldart C类脱硫灰颗粒在环流耦合提升管内稳定流动特性[J]. 化工学报, 2024, 75(4): 1485-1496.
[7]	李云璇, 刘新悦, 陈熙, 刘文, 周明月, 蓝兴英. 基于固液氧化还原靶向反应的能量存储技术：材料、器件及动力学[J]. 化工学报, 2024, 75(4): 1222-1240.
[8]	范以薇, 刘威, 李盈盈, 王培霞, 张吉松. 有机液体储氢中全氢化乙基咔唑催化脱氢研究进展[J]. 化工学报, 2024, 75(4): 1198-1208.
[9]	成文凯, 颜金钰, 王嘉骏, 冯连芳. 卧式捏合反应器及其在聚合工业中的研究进展[J]. 化工学报, 2024, 75(3): 768-781.
[10]	陈饶, 赵鑫, 陈戴欣, 姜圣坤, 廉应江, 王金波, 杨梅, 陈光文. 微反应器内甲苯连续二硝化制备二硝基甲苯[J]. 化工学报, 2024, 75(3): 867-876.
[11]	马韶阳, 徐涵卓, 张亮亮, 孙宝昌, 邹海魁, 罗勇, 初广文. 液-液非均相反应器研究进展[J]. 化工学报, 2024, 75(3): 727-742.
[12]	麻雪怡, 刘克勤, 胡激江, 姚臻. POE溶液聚合反应器内混合与反应过程的CFD研究[J]. 化工学报, 2024, 75(1): 322-337.
[13]	王婷, 王忠东, 项星宇, 何呈祥, 朱春英, 马友光, 付涛涛. 微反应器内环酯类锂电池添加剂合成研究进展[J]. 化工学报, 2024, 75(1): 95-109.
[14]	郑玉圆, 葛志伟, 韩翔宇, 王亮, 陈海生. 中高温钙基材料热化学储热的研究进展与展望[J]. 化工学报, 2023, 74(8): 3171-3192.
[15]	杨峥豪, 何臻, 常玉龙, 靳紫恒, 江霞. 生物质快速热解下行式流化床反应器研究进展[J]. 化工学报, 2023, 74(6): 2249-2263.