基于树模型和神经网络的供热一次网负荷预测

doi:10.3969/j.issn.2097-0706.2025.12.008

综合智慧能源 ›› 2025, Vol. 47 ›› Issue (12): 73-80.doi: 10.3969/j.issn.2097-0706.2025.12.008

基于树模型和神经网络的供热一次网负荷预测

闫京¹(), 蒋泽龄²(), 关宝良¹, 孟思宇¹, 王凤龙¹, 杨尚峰¹, 杨众杨¹, 熊亚选²^,^*()

1.北京天岳恒房屋经营管理有限公司，北京 100032
2.北京建筑大学供热、供燃气、通风及空调工程北京市重点实验室，北京 100044

收稿日期:2025-06-05 修回日期:2025-09-16 出版日期:2025-12-25
通讯作者: * 熊亚选（1977），男，教授，博士，从事低碳储能和供热系统精准节能方面的研究，xiongyaxuan@bucea.edu.cn。
作者简介:闫京（1974），男，高级工程师，硕士，从事城镇供热运行管理方面的工作，tyhgongnuan@163.com；
蒋泽龄（2002），女，硕士生，从事储热材料设计开发和热物性提升方面的研究，876537897@qq.com。
基金资助:
国家重点研发计划项目(2025YFE0118800)

Load prediction of primary heating networks based on tree models and neural networks

YAN Jing¹(), JIANG Zeling²(), GUAN Baoliang¹, MENG Siyu¹, WANG Fenglong¹, YANG Shangfeng¹, YANG Zhongyang¹, XIONG Yaxuan²^,^*()

1. Beijing Tianyueheng Housing Management & Administration Company Limited， Beijing 100032， China
2. Beijing Key Laboratory of Heating， Gas Supply， Ventilating and Air Conditioning Engineering， Beijing University of Civil Engineering and Architecture， Beijing 100044， China

Received:2025-06-05 Revised:2025-09-16 Published:2025-12-25
Supported by:
National Key R&D Program of China(2025YFE0118800)

摘要/Abstract

摘要：

建筑热负荷的精准预测对于优化建筑能源系统的运行、降低能源消耗及实现节能目标具有至关重要的作用。运用机器学习（ML）算法进行预测能够有效克服传统方法的局限性，显著降低传统模拟分析的计算成本并提升系统能效。采用随机森林回归（RFR）、极端随机树回归（ETR）、梯度提升回归（GBR）、极端梯度提升回归（XGBR）及多层感知机（MLP） 5种回归模型对建筑热负荷进行预测，并使用4个评价指标来评估预测精度。结果表明，ETR模型与XGBR模型的预测性能最佳。ETR模型的均方根误差（RMSE）低至97.189 4 kW，R²值达0.766 0。XGBR模型的平均绝对误差（MAE）与平均绝对百分比误差（MAPE）分别低至69.967 1 kW和4.086 0%。这2种模型均展现出较高的预测精度，为后续建筑热负荷预测研究提供了参考。

关键词: 建筑能耗, 负荷预测, 机器学习, 树模型, 神经网络, 相关系数, 建筑能源系统, 供热

Abstract:

Accurate prediction of building heating load is crucial for optimizing the operation of building energy systems， reducing energy consumption， and achieving building energy-saving goals. Using machine learning algorithms for prediction can effectively overcome the limitations of traditional prediction methods， significantly reducing the computational cost of conventional simulation analyses and enhancing system energy efficiency. Five regression machine learning models — random forest regression （RFR）， extremely randomized trees regression （ETR）， gradient boosting regression （GBR）， extreme gradient boosting regression （XGBR）， and multilayer perceptron （MLP） — were employed to predict building heating load. Four indicators were used to evaluate their prediction accuracy. The results showed that the ETR and XGBR models demonstrated the optimal predictive performance among all models. The ETR model achieved a root mean square error （RMSE） as low as 97.189 4 kW and an R² value of 0.766 0. The XGBR model achieved a mean absolute error （MAE） and a mean absolute percentage error （MAPE） as low as 69.967 1 kW and 4.086 0%， respectively. These two models achieve high predictive accuracy， providing valuable references for subsequent research on building heating load prediction.

Key words: building energy consumption, load prediction, machine learning, tree model, neural network, correlation coefficient, building energy systems, heat supply

中图分类号:

TK89

闫京, 蒋泽龄, 关宝良, 孟思宇, 王凤龙, 杨尚峰, 杨众杨, 熊亚选. 基于树模型和神经网络的供热一次网负荷预测[J]. 综合智慧能源, 2025, 47(12): 73-80.

YAN Jing, JIANG Zeling, GUAN Baoliang, MENG Siyu, WANG Fenglong, YANG Shangfeng, YANG Zhongyang, XIONG Yaxuan. Load prediction of primary heating networks based on tree models and neural networks[J]. Integrated Intelligent Energy, 2025, 47(12): 73-80.

图/表 8

表1

图 1

图2

图3

图4

表2

图5

表3

各模型的评估指标结果

模型	R_MSE/kW	M_AE/kW	M_APE/%	$R 2$
RFR	97.401 9	70.104 7	4.099 0	0.765 0
ETR	97.189 4	69.975 9	4.102 0	0.766 0
GBR	99.406 0	71.103 0	4.167 0	0.755 2
XGBR	97.800 3	69.967 1	4.086 0	0.763 1
MLP	103.872 6	73.294 7	4.271 0	0.732 7

表3

参考文献 28

[1]	杨子艺, 胡姗, 徐天昊, 等. 面向碳中和的各国建筑运行能耗与碳排放对比研究方法及应用[J]. 气候变化研究进展, 2023, 19(6): 749-760.
	YANG Ziyi, HU Shan, XU Tianhao, et al. Method and application of global building operation energy use and carbon emissions comparison in the context of carbon neutrality[J]. Climate Change Research, 2023, 19(6): 749-760.
[2]	LIU R X, KUANG J, GONG Q, et al. Principal component regression analysis with SPSS[J]. Computer Methods and Programs in Biomedicine, 2003, 71(2): 141-147. doi: 10.1016/s0169-2607(02)00058-5 pmid: 12758135
[3]	吕岩, 潘毅群, 刘海静, 等. 冷热负荷预测在区域供能项目中的应用: 以上海西虹桥1号能源站为例[J]. 全球能源互联网, 2021, 4(2): 197-203.
	LYU Yan, PAN Yiqun, LIU Haijing, et al. Application of cooling and heating load prediction for district energy supply: A case in west Hongqiao 1^# energy station[J]. Journal of Global Energy Interconnection, 2021, 4(2): 197-203.
[4]	庄惟敏, 刘加平, 王建国, 等. 建筑碳中和的关键前沿基础科学问题[J]. 中国科学基金, 2023, 37(3): 348-352.
	ZHUANG Weimin, LIU Jiaping, WANG Jianguo, et al. Key frontier basic scientific issues in building carbon neutrality[J]. Bulletin of National Natural Science Foundation of China, 2023, 37(3): 348-352.
[5]	LU C J, LI S H, LU Z J. Building energy prediction using artificial neural networks: A literature survey[J]. Energy and Buildings, 2022, 262: 111718. doi: 10.1016/j.enbuild.2021.111718
[6]	ZHANG L, WEN J, LI Y F, et al. A review of machine learning in building load prediction[J]. Applied Energy, 2021, 285: 116452. doi: 10.1016/j.apenergy.2021.116452
[7]	FATHI S, SRINIVASAN R, FENNER A, et al. Machine learning applications in urban building energy performance forecasting: A systematic review[J]. Renewable and Sustainable Energy Reviews, 2020, 133: 110287. doi: 10.1016/j.rser.2020.110287
[8]	徐聪, 胡永锋, 张爱平, 等. 基于特征筛选的综合能源系统多元负荷日前-日内预测[J]. 综合智慧能源, 2024, 46(3): 45-53. doi: 10.3969/j.issn.2097-0706.2024.03.006
	XU Cong, HU Yongfeng, ZHANG Aiping, et al. Multi-load day-ahead and intra-day forecasting for integrated energy systems based on feature screening[J]. Integrated Intelligent Energy, 2024, 46(3): 45-53. doi: 10.3969/j.issn.2097-0706.2024.03.006
[9]	杨澜倩, 郭锦敏, 田慧丽, 等. 基于CNN-LSTM-Self attention的园区负荷多尺度预测研究[J]. 综合智慧能源, 2025, 47(2): 79-87. doi: 10.3969/j.issn.2097-0706.2025.02.008
	YANG Lanqian, GUO Jinmin, TIAN Huili, et al. Research on multi-scale load prediction in parks based on CNN-LSTM-Self attention[J]. Integrated Intelligent Energy, 2025, 47(2): 79-87. doi: 10.3969/j.issn.2097-0706.2025.02.008
[10]	LU C J, LI S H, REDDY PENAKA S, et al. Automated machine learning-based framework of heating and cooling load prediction for quick residential building design[J]. Energy, 2023, 274: 127334. doi: 10.1016/j.energy.2023.127334
[11]	WEI Z Q, ZHANG T W, YUE B, et al. Prediction of residential district heating load based on machine learning: A case study[J]. Energy, 2021, 231: 120950. doi: 10.1016/j.energy.2021.120950
[12]	GAO T F, HAN X, WANG J, et al. Enhancing building energy efficiency: An integrated approach to predicting heating and cooling loads using machine learning and optimization algorithms[J]. Journal of Building Engineering, 2024, 98: 110759. doi: 10.1016/j.jobe.2024.110759
[13]	GUO J X, YUN S N, MENG Y, et al. Prediction of heating and cooling loads based on light gradient boosting machine algorithms[J]. Building and Environment, 2023, 236: 110252. doi: 10.1016/j.buildenv.2023.110252
[14]	BREIMAN L. Random forests[J]. Machine Learning, 2001, 45(1): 5-32. doi: 10.1023/A:1010933404324
[15]	JIANG R, TANG W W, WU X B, et al. A random forest approach to the detection of epistatic interactions in case-control studies[J]. BMC Bioinformatics, 2009, 10(Suppl 1): S65. doi: 10.1186/1471-2105-10-S1-S65
[16]	AHMAD M W, MOURSHED M, REZGUI Y. Tree-based ensemble methods for predicting PV power generation and their comparison with support vector regression[J]. Energy, 2018, 164: 465-474. doi: 10.1016/j.energy.2018.08.207
[17]	MARANI A, NEHDI M L. Machine learning prediction of compressive strength for phase change materials integrated cementitious composites[J]. Construction and Building Materials, 2020, 265: 120286. doi: 10.1016/j.conbuildmat.2020.120286
[18]	JOHN V, LIU Z, GUO C Z, et al. Real-time lane estimation using deep features and extra trees regression[M]//Image and Video Technology. Cham: Springer International Publishing, 2016: 721-733.
[19]	AHMAD M W, REYNOLDS J, REZGUI Y. Predictive modelling for solar thermal energy systems: A comparison of support vector regression, random forest, extra trees and regression trees[J]. Journal of Cleaner Production, 2018, 203: 810-821. doi: 10.1016/j.jclepro.2018.08.207
[20]	ALAM M S, AL-ISMAIL F S, HOSSAIN M S, et al. Ensemble machine-learning models for accurate prediction of solar irradiation in Bangladesh[J]. Processes, 2023, 11(3): 908. doi: 10.3390/pr11030908
[21]	GUELMAN L. Gradient boosting trees for auto insurance loss cost modeling and prediction[J]. Expert Systems with Applications, 2012, 39(3): 3659-3667. doi: 10.1016/j.eswa.2011.09.058
[22]	CHANG Y C, CHANG K H, WU G J. Application of eXtreme gradient boosting trees in the construction of credit risk assessment models for financial institutions[J]. Applied Soft Computing, 2018, 73: 914-920. doi: 10.1016/j.asoc.2018.09.029
[23]	YUDANTAKA K, KIM J S, SONG H, et al. Dual deep learning networks based load forecasting with partial real-time information and its application to system marginal price prediction[J]. Energies, 2019, 13(1): 1-15. doi: 10.3390/en13010001
[24]	GARDNER M W, DORLING S R. Artificial neural networks (the multilayer perceptron): A review of applications in the atmospheric sciences[J]. Atmospheric Environment, 1998, 32(14/15): 2627-2636. doi: 10.1016/S1352-2310(97)00447-0
[25]	XIE Y H, UEDA Y, SUGIYAMA M, et al. A two-stage short-term load forecasting method using long short-term memory and multilayer perceptron[J]. Energies, 2021, 14(18): 5873. doi: 10.3390/en14185873
[26]	杨丽洁, 邓振宇, 陈作双, 等. 基于MSCNN-BiGRU-MLP模型的公共建筑非侵入式负荷辨识[J]. 综合智慧能源, 2025, 47(3): 23-31. doi: 10.3969/j.issn.2097-0706.2025.03.003
	YANG Lijie, DENG Zhenyu, CHEN Zuoshuang, et al. Non-intrusive load identification for public buildings based on MSCNN-BiGRU-MLP model[J]. Integrated Intelligent Energy, 2025, 47(3): 23-31. doi: 10.3969/j.issn.2097-0706.2025.03.003
[27]	刘艺娴, 王玉彬, 杨强. 基于门控图神经网络的高容错配电网状态估计方法[J]. 综合智慧能源, 2023, 45(6): 1-8. doi: 10.3969/j.issn.2097-0706.2023.06.001
	YLIU Yixian, WANG Yubin, YANG Qiang. High fault-tolerant distribution network state estimation method based on gated graph neural network[J]. Integrated Intelligent Energy, 2023, 45(6): 1-8. doi: 10.3969/j.issn.2097-0706.2023.06.001
[28]	ATTALI J G, PAGÈS G. Approximations of functions by a multilayer perceptron: A new approach[J]. Neural Networks, 1997, 10(6): 1069-1081. pmid: 12662500

参数	室外干球温度/℃	室外风速/(m·s^-1)	室外相对湿度/%	太阳辐照度/(W·m^-2)	热泵供水温度/℃	热泵回水温度/℃	热负荷/kW
最小值	-5.78	30.52	0.04	0	34.28	31.22	1 187.325 9
最大值	13.07	64.25	10.84	771.00	41.98	35.93	2 286.033 7
平均值	4.80	42.76	1.33	87.29	38.23	33.08	1 760.960 7

基于树模型和神经网络的供热一次网负荷预测

Load prediction of primary heating networks based on tree models and neural networks

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 8

参考文献 28

相关文章 15

编辑推荐

Metrics

本文评价

模型	参数
RFR	决策树数量=400；最小分割样本数=2；叶节点最小样本数=1；最大分割特征数='log2'；树的最大深度=14
ETR	决策树数量=400；最小分割样本数=3；叶节点最小样本数=1；最大分割特征数=None；树的最大深度=13
GBR	决策树数量=400；最小分割样本数=2；叶节点最小样本数=1；最大分割特征数='log2'；树的最大深度=10；学习率=0.01
XGBR	子采样率=0.8；决策树数量=100；最小叶子权重=5；树的最大深度=10；学习率=0.05；分割阈值=0.1；特征采样比例=0.5

[1]	黄子书, 蔡晔, 孙溶佐, 谭玉东. 电力系统高危N-k故障的高危断面辨识与保护配置[J]. 综合智慧能源, 2025, 47(9): 1-9.
[2]	李祯, 杨国华, 张元曦, 马鑫, 杨娜, 刘浩睿, 马龙腾. 基于模态二次分解和OOA-CNN-BiLSTM-Attention的光伏发电功率组合预测[J]. 综合智慧能源, 2025, 47(9): 28-37.
[3]	皇甫陈萌, 阮贺彬, 徐俊俊. 融合CEEMDAN-CNN-LSTM的风电机组多气象场景功率回归预测[J]. 综合智慧能源, 2025, 47(9): 38-50.
[4]	窦翔, 李卓群, 张哲, 温鑫, 赵勃, 韩燕, 仲声. 基于CNN-BiLSTM-RF-KDE的综合能源系统负荷预测[J]. 综合智慧能源, 2025, 47(9): 60-70.
[5]	殷林飞, 张依玲. 结合变分模态分解与三重卷积神经网络的光伏出力预测[J]. 综合智慧能源, 2025, 47(6): 11-19.
[6]	程先龙, 张杰, 李思莹, 杨翼霞, 杨翠飞. 基于VMD-BP-BiLSTM的短期风电功率预测[J]. 综合智慧能源, 2025, 47(6): 20-29.
[7]	湛国华, 张先勇, 魏圣莹, 张孝顺, 李丽. 基于T-Graphormer的电网碳排放因子预测方法[J]. 综合智慧能源, 2025, 47(6): 30-36.
[8]	蒋剑, 杜董生, 苏林. 基于改进HHO-LSTM-Self-Attention的质子交换膜燃料电池剩余使用寿命预测[J]. 综合智慧能源, 2025, 47(6): 47-56.
[9]	班逢春, 陈萧凤, 黄志甲. 基于PSO-BP神经网络的住宅光伏发电预测模型[J]. 综合智慧能源, 2025, 47(6): 85-93.
[10]	黄琳杰, 谢芷珊, 廖永兴, 殷林飞. 考虑疲劳损伤的多级前馈神经网络-序列二次规划的风电场降噪优化[J]. 综合智慧能源, 2025, 47(4): 23-32.
[11]	陆汉东, 方明, 刘刚刚, 周妍. 基于传播模型与神经网络的输变电工程造价分析与预测方法研究[J]. 综合智慧能源, 2025, 47(4): 33-40.
[12]	殷林飞, 张依玲. 基于多重卷积组合大模型的光伏出力预测[J]. 综合智慧能源, 2025, 47(4): 63-72.
[13]	黄问铉, 曾浩政, 林奕津, 殷林飞. 基于数据驱动方法的磁性元件损耗研究[J]. 综合智慧能源, 2025, 47(4): 73-84.
[14]	杨丽洁, 邓振宇, 陈作双, 黄超, 江美慧, 朱虹谕. 基于MSCNN-BiGRU-MLP模型的公共建筑非侵入式负荷辨识[J]. 综合智慧能源, 2025, 47(3): 23-31.
[15]	张冬冬, 李芳凝, 刘天皓. 新型电力系统负荷预测关键技术及多元场景应用[J]. 综合智慧能源, 2025, 47(3): 47-61.