基于神经网络与遗传算法的多热源冷板结构优化设计

doi:10.3969/j.issn.1008-0198.2026.05.010

湖南电力 ›› 2026, Vol. 46 ›› Issue (5): 76-84.doi: 10.3969/j.issn.1008-0198.2026.05.010

基于神经网络与遗传算法的多热源冷板结构优化设计

张逸飞, 庄煜彬, 崔盼, 刘志春

华中科技大学能源与动力工程学院,武汉 430074

收稿日期:2025-12-29 修回日期:2026-01-27 发布日期:2026-11-06
通信作者: 刘志春（1976）,男,教授,博士生导师,主要研究方向为强化传热理论与技术、电子器体散热等。
作者简介:张逸飞（2001）,男,硕士研究生,主要研究方向为多热源电子器件强化换热。
基金资助:
国家重点研发计划项目（2022YFB4003801）

Structural Optimization of Multi-Heat-Source Cold Plates Using Neural Network and Genetic Algorithm

ZHANG Yifei, ZHUANG Yubin, CUI Pan, LIU Zhichun

School of Energy and Power Engineering, Huazhong University of Science and Technology, Wuhan 430074, China

Received:2025-12-29 Revised:2026-01-27 Published:2026-11-06

摘要/Abstract

摘要： 针对多热源液冷板结构优化问题,提出一种结合神经网络与遗传算法的直肋冷板流道多区域独立调控优化方法,采用流道翅片按热源分布分区域独立控制策略,以热源最高温度及进出口压降为性能指标,阐述整体流道控制方案的局限性,通过神经网络与遗传算法构建双优化目标。计算流体力学模型计算结果显示：压降最小目标使流阻减小47.4%,比整体优化方案降低34.9%;综合性能目标使流阻减小28%,比整体优化方案降低12.2%;同时关键区域温度降低0.54 ℃,且各热源温度均处于合理范围。该方法为多热源液冷板设计提供参考,可根据热力与水力目标权重灵活制定策略,有效提升设计效率与综合性能。

关键词: 多热源冷板, 神经网络, 遗传算法, 结构优化, 参数研究

Abstract: For the structural optimization of multi-heat-source liquid-cooled plates, a multi-zone independent regulation and optimization method for straight rib cold plate flow channels is proposed, combining neural networks and genetic algorithms. Using the maximum temperature of heat sources and the inlet-outlet pressure difference as performance indicators, the limitations of conventional overall channel control are highlighted. Two optimization objectives are formulated and solved by combining via neural networks and genetic algorithms. The results show that minimizing the pressure difference reduces flow resistance by 47.4%, which is 34.9% lower than the result obtained via the conventional full-domain optimization. Under the comprehensive performance objective, flow resistance is reduced by 28% (12.2% lower than the overall optimization), while the temperature in critical regions decreases by 0.54 ℃, with all heat source temperatures maintained within reasonable ranges. This method provides a scalable framework for efficient design of multi-heat-source liquid-cooled cold plates, enabling flexible strategy formulation based on the weighting of thermal and hydraulic objectives, thereby effectively improving design efficiency and overall performance.

Key words: multi-heat source cold plate, neural network, genetic algorithm, structural optimization, parametric study

中图分类号:

TK124

张逸飞, 庄煜彬, 崔盼, 刘志春. 基于神经网络与遗传算法的多热源冷板结构优化设计[J]. 湖南电力, 2026, 46(5): 76-84.

ZHANG Yifei, ZHUANG Yubin, CUI Pan, LIU Zhichun. Structural Optimization of Multi-Heat-Source Cold Plates Using Neural Network and Genetic Algorithm[J]. Hunan Electric Power, 2026, 46(5): 76-84.

参考文献

[1] 李广义,张俊洪,高键鑫. 大功率电力电子器件散热研究综述[J]. 兵器装备工程学报,2020,41(11):8-14.
[2] Bahun I,Sunde V,Jakopovic Z,Estimation of insulated-gate bipolar transistor operating temperature simulation and experiment[J]. Journal of Power Electronics,2013,13(4):729-736.
[3] 李建. 多样化算力对服务器的散热挑战分析[J]. 信息通信技术与政策,2024,50(2):46-54.
[4] Chen X G,Hu C D,Liang J,et al.Efficient two-phase liquid cooling and optimization technology on the chip side based on high-power data centers:Summary and prospects[J]. Renewable and Sustainable Energy Reviews,2025,223:115994.
[5] Huang H Z,Bu C G.Design and verification of thermal control system of communication satellite[J]. Aerospace,2024,11(10):803.
[6] 杨双根,金开军,朱春玲. 液冷组件热设计[J]. 安徽工程科技学院学报(自然科学版),2004,19(4):73-76.
[7] 束瑛. 微通道冷板导热性优化设计[J]. 电子机械工程,2009,25(2):8-15.
[8] 夏侯国伟,王当,刘业鹏. IGBT功率模块冷却技术的综述[J]. 昆明理工大学学报(自然科学版),2017,42(1):63-67,84.
[9] 区嘉洁. 清洁燃气客车发动机舱多场耦合强化散热原理研究及其应用[D]. 广州:华南理工大学. 2018.
[10] 张荣. 热管冷板在传导冷却LRM模块上的应用[J]. 机械与电子,2017,35(6):240-244.
[11] 魏涛,方晓鹏,钱吉裕. 一种基于场协同原理的强化液冷冷板研究[J]. 现代雷达,2018,40(12):87-90.
[12] Shi B,Srivastava A.Cooling of 3D-IC using non-uniform micro-channels and sensor based dynamic thermal management[C]//2011 49th Annual Allerton Conference on Communication Control and Computing(Allerton). Monticello,LL. IEEE,2011:1400-1407.
[13] Zhang M K,Zhang X P,Guo L N,et al.Flow and thermal modeling of liquid metal in expanded microchannel heat sink[J]. Frontiers in Energy,2023,17(6):796-810.
[14] Liu F Q,Jia L,Zhang J X,et al.Numerical study on heat transfer efficiency and inter-layer stress of microchannel heat sinks with different geometries[J]. Energies. 2024,17(20):5076.
[15] Ansari D,Raza W,Jeong J H,et al.A novel hybrid-composite microchannel heat sink for extreme hotspot mitigation[J]. International Journal of Thermal Sciences,2025,208:109473.
[16] Alihosseini Y,Oghabneshin Y,Bari A R,et al.Oblique microchannel merged with circle micro pin-fin as a novel hybrid heat sink for cooling of electronic devices[J]. Case Studies in Thermal Engineering,2024,53:103888.
[17] Lee H,Kang M,Jung K W,et al.An artificial neural network model for predicting frictional pressure drop in micro-pin fin heat sink[J]. Applied Thermal Engineering,2021,117012.
[18] Feng C Y,Zhang P,Wang D W,et al.Hybrid neural network based multi-objective optimal design of hybrid pin-fin microchannel heatsink for integrated microsystems[J]. International Communications in Heat and Mass Transfer,2024,159:108137.
[19] He Z X,Yu Q H,Ye J D,et al.Optimization of plate-fin heat exchanger performance for heat dissipation of thermoelectric cooler[J]. Case Studies in Thermal Engineering,2024,53:103953.
[20] Deb K,Pratap A,Agarwal S,et al.A fast and elitist multi-objective genetic algorithm:NSGA-II[J]. IEEE Transactions on Evolutionary Computation,2002,6(2):182-197.

基于神经网络与遗传算法的多热源冷板结构优化设计

Structural Optimization of Multi-Heat-Source Cold Plates Using Neural Network and Genetic Algorithm

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 15

编辑推荐

Metrics

本文评价

[1]	谢捷敏, 黄波, 唐传奇, 张弘强. 结合小波变换与多尺度卷积时序融合网络的轴流泵声信号故障诊断[J]. 湖南电力, 2026, 46(5): 135-141.
[2]	贺继林, 陈伟, 杨淼, 殷鹏. 面向配电网电杆组立施工的钻孔立杆机自适应轨迹跟踪控制研究[J]. 湖南电力, 2026, 46(4): 79-88.
[3]	张锐, 胡旭光, 杜杭远, 赵建策. 基于多粒度Koopman模态图感知结构的电缆故障诊断方法[J]. 湖南电力, 2026, 46(4): 113-120.
[4]	吴冲, 何洪浩, 杨苡清, 徐俊, 朱光明, 向军. 基于WOA-VMD-CNN-BiGRU的锅炉水冷壁壁温预测模型研究[J]. 湖南电力, 2026, 46(2): 7-14.
[5]	詹子仪, 黄延凯, 喻鑫, 周子健, 于敦喜, 徐明厚. 用于复杂煤灰熔融特性预测的加权耦合神经网络模型[J]. 湖南电力, 2025, 45(6): 68-75.
[6]	李文澜, 鲁强. 基于主题-神经网络的电网企业高质量发展规划评价[J]. 湖南电力, 2025, 45(5): 104-111.
[7]	孙晓敏, 温志超, 谢浩南, 吴宏曜, 宋冬然. 基于改进遗传算法和分支定界法的行波检测装置双层优化配置方法[J]. 湖南电力, 2025, 45(4): 83-89.
[8]	张亦可, 张军, 莫剑, 张学武, 张振. 多设备融合的水工建筑物水下检测技术[J]. 湖南电力, 2025, 45(3): 35-41.
[9]	张大弛, 杨士伟, 丰顺强, 姚狄, 辛昊阔, 肖白. 基于MTF-PSO-CNN的电能质量扰动分类方法[J]. 湖南电力, 2025, 45(3): 103-111.
[10]	李勃铖, 杨昊彦, 吴坚, 刘雁灿, 刘华洲, 陈彬. 基于RBF神经网络算法的复合绝缘子红外测温修正方法[J]. 湖南电力, 2025, 45(3): 135-140.
[11]	耿萌萌, 范茂松, 官亦标, 魏斌. 基于运行数据的电化学储能电池SOH评估[J]. 湖南电力, 2025, 45(1): 8-13.
[12]	于志诚, 穆士才, 梁晔, 李镓辰, 林华, 陈己宸, 金鑫. 基于K-means聚类算法和BP神经网络的代理购电量预测模型研究[J]. 湖南电力, 2025, 45(1): 68-72.
[13]	梁海维, 王阳光, 邓小亮, 刘静, 文明, 于宗超, 李文英. 基于分段预测及天气相似日选择的区域电网短期负荷预测方法[J]. 湖南电力, 2024, 44(5): 109-116.
[14]	文思伦, 张楚岩, 刁明光, 周振安, 徐惠勇, 刘慧芳. 基于LSTM和ArcGIS的区域输电网污秽状态智能评估方法[J]. 湖南电力, 2024, 44(3): 55-63.
[15]	杨芳僚, 黄鑫, 谭鸿志, 闵琦, 祝视, 燕磊. 基于信息融合与一维卷积神经网络的光伏电站传感器健康状态评估方法[J]. 湖南电力, 2024, 44(3): 105-113.