基于机器学习与粒子群算法的微结构辐射冷却性能优化研究

doi:10.11949/0438-1157.20250667

摘要/Abstract

摘要：

为构建具有高发射率的表面微结构，提高其辐射制冷性能，针对一种具有四棱台周期性特征的辐射制冷微结构，开展了其大气窗口发射率的优化研究。首先，基于时域有限差分（FDTD）方法，计算了不同几何参数条件下的光谱特性；然后，基于大量数据，建立了利用神经网络方法的发射率预测机器学习模型；最后，结合自适应粒子群优化算法（APSO）实现了发射性能优化，获得最佳几何参数组合，并分析了不同结构参数对大气窗口平均发射率的影响。结果发现，利用FDTD计算结果和神经网络算法的机器学习模型，可以实现此类结构发射率的高精度预测。不同几何参数对发射性能的影响不同，最佳参数组合下，此类微结构的大气窗口平均发射率接近于1，具有很强的辐射制冷能力。

关键词: 制冷性能, 辐射, 微尺度, 神经网络, 结构优化

Abstract:

To construct surface microstructures with high emissivity and improve their radiative cooling performance, this study investigated the optimization of the atmospheric window emissivity of a radiative cooling microstructure with periodic tetrahedral features. Firstly, based on the finite-difference time-domain (FDTD) method, the spectral characteristics under different geometric parameter conditions are calculated. Then, based on a large amount of data, a machine-learning model for emissivity prediction using the neural network method is established. Finally, by combining with the adaptive particle swarm optimization (APSO) algorithm, the emission performance is optimized, the optimal combination of geometric parameters is obtained, and the influence of different structural parameters on the average emissivity in the atmospheric window is analyzed. The results show that the machine-learning model based on the FDTD calculation results and neural network algorithms can achieve high-precision emissivity prediction for such structures. Different geometric parameters have different impacts on the emissivity. Under the optimal combination of structural parameters, the average emissivity of this type of microstructure in the atmospheric window is close to 1, indicating strong radiative cooling ability.

Key words: cooling performance, radiation, microscale, neural networks, structure optimization

中图分类号:

TK 121

刘博, 黄昊辉, 李奇蕴, 孙晨贸, 冯杰. 基于机器学习与粒子群算法的微结构辐射冷却性能优化研究[J]. 化工学报, 2025, 76(12): 6626-6632.

Bo LIU, Haohui HUANG, Qiyun LI, Chenmao SUN, Jie FENG. Optimization of radiative cooling performance of microstructure based on machine learning model and particle swarm algorithm[J]. CIESC Journal, 2025, 76(12): 6626-6632.

图/表 8

图1 特征微结构和几何参数

Fig.1 Characteristic structure and geometric parameters

图2 本文采用FDTD方法的准确性验证

Fig.2 Verification of adopted FDTD method

图3 神经网络模型示意图

Fig.3 Schematic diagram of neural network model

图4 机器学习模型预测结果的响应图

Fig.4 response diagram of prediction results of machine learning model

表1 机器学习模型预测结果的统计指标

Table 1 Statistical indicators of prediction results of machine learning model

统计指标	数值
RMSE	0.011314
MSE	0.000128
R²	0.99841
MAE	0.0083269

图5 机器学习模型与APSO算法的联合优化流程

Fig.5 Joint optimization process of machine learning model and APSO algorithm

图6 最佳辐射制冷结构的光谱发射特性

Fig.6 Spectral emission characteristics of the optimal radiation refrigeration structure

图7 几何参数对微结构大气窗口平均发射率的影响

Fig.7 Effect of geometric parameters on average emissivity of microstructure atmospheric window

参考文献 31

[1]	Lee M, Kim G, Jung Y, et al. Photonic structures in radiative cooling[J]. Light: Science & Applications, 2023, 12: 134.
[2]	刘冠水. 辐射制冷技术在高铁站房的节能应用研究[J]. 制冷技术, 2024, 44(6): 75-79, 86.
	Liu G S. Study on energy-saving application of radiative cooling technology in high-speed railway station[J]. Chinese Journal of Refrigeration Technology, 2024, 44(6): 75-79, 86.
[3]	徐第开, 盛茗峰, 杨荣贵, 等.天空辐射制冷规模化应用对我国建筑的减碳作用研究[J]. 制冷学报, 2023, 44(6): 13-21, 28.
	Xu D K, Sheng M F, Yang R G, et al. Large-scale application of radiative sky cooling in buildings for carbon emission reduction[J]. Journal of Refrigeration. 2023, 44(6): 13-21, 28.
[4]	文凯, 王程远, 王晓坡, 等. 耦合天空辐射制冷的数据中心自然冷却方案分析[J]. 工程热物理学报, 2024, 45(5): 1248-1254.
	Wen K, Wang C Y, Wang X P, et al. Analysis of free cooling scheme coupled with radiative sky cooling in data center[J]. Journal Of Engineering Thermophysics, 2024, 45(5): 1248-1254.
[5]	曹雄金, 王艳, 王磊, 等. 相变材料与辐射制冷材料耦合技术的研究进展[J]. 化工新型材料, 2025, 53(3): 58-64.
	Cao X J, Wang Y, Wang L, et al. Research progress on coupling technology of phase change materials and radiation cooling material[J]. New Chemical Materials, 2025, 53(3): 58-64.
[6]	Lin K T, Han J H, Li K, et al. Radiative cooling: fundamental physics, atmospheric influences, materials and structural engineering, applications and beyond[J]. Nano Energy, 2021, 80: 105517.
[7]	Sheng M F, Pan H D, Xu D K, et al. Characterization and performance enhancement of radiative cooling on circular surfaces[J]. Renewable and Sustainable Energy Reviews, 2023, 188: 113782.
[8]	Ahmed S, Li Z P, Javed M S, et al. A review on the integration of radiative cooling and solar energy harvesting[J]. Materials Today Energy, 2021, 21: 100776.
[9]	Yu X X, Chen C. Coupling spectral-dependent radiative cooling with building energy simulation[J]. Building and Environment, 2021, 197: 107841.
[10]	Zeyghami M, Goswami D Y, Stefanakos E. A review of clear sky radiative cooling developments and applications in renewable power systems and passive building cooling[J]. Solar Energy Materials and Solar Cells, 2018, 178: 115-128.
[11]	Yang X F, Wang S T, Zhang D, et al. Theoretical investigation of two ideal radiative cooling materials and radiative sky water-cooling module with ideal selective radiative materials[J]. Energy and Buildings, 2024, 323: 114768.
[12]	Li B, Cao B Y, Song R C, et al. Low-cost and scalable sub-ambient radiative cooling porous films[J]. Journal of Photonics for Energy, 2023, 13(1): 015501.
[13]	Chae D, Son S, Lim H, et al. Scalable and paint-format microparticle-polymer composite enabling high-performance daytime radiative cooling[J]. Materials Today Physics, 2021, 18: 100389.
[14]	Bijarniya J P, Sarkar J, Maiti P. Review on passive daytime radiative cooling: fundamentals, recent researches, challenges and opportunities[J]. Renewable and Sustainable Energy Reviews, 2020, 133: 110263.
[15]	Ma H C, Yao K Q, Dou S L, et al. Multilayered SiO₂/Si₃N₄ photonic emitter to achieve high-performance all-day radiative cooling[J]. Solar Energy Materials and Solar Cells, 2020, 212: 110584.
[16]	Chen J, Li X, Chen Y T, et al. Temperature self-adaptive ultra-thin solar absorber based on optimization algorithm[J]. Photonics, 2023, 10(5): 546.
[17]	刘立君, 李龙飞, 毛前军. 基于遗传算法优化光栅辐射特性[J]. 工程热物理学报, 2016, 37(7): 1538-1543.
	Liu L J, Li L F, Mao Q J. Radiative property optimization of gratings based on genetic algorithm[J]. Journal of Engineering Thermophysics, 2016, 37(7): 1538-1543.
[18]	Han T, Zhou Z H, Du Y H, et al. Advances in radiative sky cooling based on the promising electrospinning[J]. Renewable and Sustainable Energy Reviews, 2024, 200: 114533.
[19]	王林茜, 陈娟, 牟春晖. 基于稀疏采样的FDTD/TDPO混合优化算法[J]. 电波科学学报, 2024, 39(5): 846-851.
	Wang L X, Chen J, Mou C H. Hybrid optimization algorithm of FDTD/TDPO based on sparse sampling[J]. Chinese Journal of Radio Science, 2024, 39(5): 846-851.
[20]	Bass S F, Palmer A M, Schab K R, et al. Conversion matrix method of moments for time-varying electromagnetic analysis[J]. IEEE Transactions on Antennas and Propagation, 2022, 70(8): 6763-6774.
[21]	邓祥文, 伍力源, 赵锐, 等. 机器学习在光电子能谱中的应用及展望[J]. 物理学报, 2024, 73(21): 64-84.
	Deng X W, Wu L Y, Zhao R, et al. Application and prospect of machine learning in photo electron spectroscopy[J]. Acta Physica Sinica, 2024, 73(21): 64-84.
[22]	Sun S L, Cao Z H, Zhu H, et al. A survey of optimization methods from a machine learning perspective[J]. IEEE Transactions on Cybernetics, 2020, 50(8): 3668-3681.
[23]	Narvaez G, Giraldo L F, Bressan M, et al. Machine learning for site-adaptation and solar radiation forecasting[J]. Renewable Energy, 2021, 167: 333-342.
[24]	杜佳欣, 王富强, 张鑫平, 等. 机器学习辅助辐射特性定向调控优化设计研究与应用综述[J]. 东北电力大学学报, 2024, 44(6): 63-73.
	Du J X, Wang F Q, Zhang X P, et al. A review on the research and application of machine learning assisted optimal design for directional regulation of radiation properties[J]. Journal of Northeast Electric Power University, 2024, 44(6): 63-73.
[25]	Kim S, Shang W J, Moon S, et al. High-performance transparent radiative cooler designed by quantum computing[J]. ACS Energy Letters, 2022, 7(12): 4134-4141.
[26]	Pan Q H, Zhou S H, Chen S N, et al. Deep learning-based inverse design optimization of efficient multilayer thermal emitters in the near-infrared broad spectrum[J]. Optics Express, 2023, 31(15): 23944.
[27]	汪丽旭, 韩玉阁. 微结构热辐射器光谱特性的影响因素研究[J]. 科学技术与工程, 2016, 16(11): 180-184.
	Wang L X, Han Y G. Research on affecting factors of spectral characteristics of radiator with microstructure[J]. Science Technology and Engineering, 2016, 16(11): 180-184.
[28]	廖涂威, 鞠生宏, 赵长颖. 机器学习加速多层双曲材料近场热调控器设计[J]. 工程热物理学报, 2025, 46(5): 1606-1612.
	Liao T W, Ju S H, Zhao C Y. Review of surrogate gradient methods in spiking deep learning[J]. Journal of Engineering Thermophysics, 2025, 46(5): 1606-1612.
[29]	Edwards C. Neural networks learn to speed up simulations[J]. Communications of the ACM, 2022, 65(5): 27-29.
[30]	Rizqi Z U, Chou S Y, Yu T H. Green energy mix modeling under supply uncertainty: hybrid system dynamics and adaptive PSO approach[J]. Applied Energy, 2023, 349: 121643.
[31]	邱千里, 章晋国, 周东劼, 等. 基于多目标粒子群优化算法设计的双波段窄带热辐射器[J]. 红外与毫米波学报, 2025, 44(1): 11-16.
	Qiu Q L, Zhang J G, Zhou D J, et al. Dual-band narrowband thermal emitter designed based on multi objective particle swarm optimization algorithm[J]. Journal of Infrared Millimeter Waves, 2025, 44(1): 11-16.