基于Transformer多尺度融合网络的暖通空调能耗预测模型

于水; 韩府宏; 罗宇晨; 孙圣坤

doi:10.19912/j.0254-0096.tynxb.2024-1746

PDF(3488 KB)

太阳能学报 ›› 2026, Vol. 47 ›› Issue (2) : 300-309. DOI: 10.19912/j.0254-0096.tynxb.2024-1746

基于Transformer多尺度融合网络的暖通空调能耗预测模型

于水, 韩府宏, 罗宇晨, 孙圣坤

作者信息 +

ENERGY CONSUMPTION PREDICTION MODEL FOR HEATING VENTILATION AIR CONDITIONING BASED ON TRANSFORMER MULTI-SCALE FUSION NETWORK

Yu Shui, Han Fuhong, Luo Yuchen, Sun Shengkun

Author information +

文章历史 +

摘要

提出一种基于Transformer的多尺度融合网络模型,用于预测建筑暖通空调的能耗。通过引入多尺度金字塔模块与时间卷积网络结构,该模型能够有效捕捉时序特征的局部与整体信息,从而提高预测的准确性。实验结果表明,该模型在预测性能上优于传统的单一模型,均方根误差（RMSE）与平均绝对误差（MAE）均显著降低,决定系数（R²）达到0.9826。该模型可为建筑能耗管理提供一种高效且准确的预测工具,有助于实现更高效的建筑能源管理与节能策略。

Abstract

As the proportion of heating venlt lation air conditioning（HVAC） energy consumption in the overall energy consumption of buildings continues to increase, improving energy consumption prediction capability is an important technical measure for enhancing the fine management of building energy consumption and achieving carbon neutrality in the building sector. This paper proposes a Transformer-based multi-scale fusion network model for predicting HVAC energy consumption in buildings. By introducing a multi-scale pyramid module and a temporal convolutional network structure, the model effectively captures both local and global temporal features, thereby improving prediction accuracy. Experimental results show that the model outperforms traditional single models in prediction performance, with a significant reduction in root mean square error （RMSE） and mean absolute error （MAE）, and a determination coefficient （R²） of 0.9826. This model provides an efficient and accurate prediction tool for building energy consumption management and contributes to more efficient building energy management and energy-saving strategies.

导出引用

于水, 韩府宏, 罗宇晨, 孙圣坤. 基于Transformer多尺度融合网络的暖通空调能耗预测模型[J]. 太阳能学报. 2026, 47(2): 300-309 https://doi.org/10.19912/j.0254-0096.tynxb.2024-1746

Yu Shui, Han Fuhong, Luo Yuchen, Sun Shengkun. ENERGY CONSUMPTION PREDICTION MODEL FOR HEATING VENTILATION AIR CONDITIONING BASED ON TRANSFORMER MULTI-SCALE FUSION NETWORK[J]. Acta Energiae Solaris Sinica. 2026, 47(2): 300-309 https://doi.org/10.19912/j.0254-0096.tynxb.2024-1746

中图分类号： TU201.4

参考文献

[1] 张时聪, 王珂, 徐伟. 低碳、近零碳、零碳公共建筑碳排放控制指标研究[J]. 建筑科学, 2023, 39(2): 1-10, 35.
ZHANG S C, WANG K, XU W.Research on carbon emission control index of low-carbon, nearly zero-carbon and zero-carbon public buildings[J]. Building science, 2023, 39(2): 1-10, 35.
[2] 宋梦, 高赐威, 苏卫华. 面向需求响应应用的空调负荷建模及控制[J]. 电力系统自动化, 2016, 40(14): 158-167.
SONG M, GAO C W, SU W H.Modeling and controlling of air-conditioning load for demand response applications[J]. Automation of electric power systems, 2016, 40(14): 158-167.
[3] 许馨尹, 李红莲, 杨柳, 等. 气候变化下的建筑能耗预测[J]. 太阳能学报, 2018, 39(5): 1359-1366.
XU X Y, LI H L, YANG L, et al.Prediction of future building energy consumption under climate change[J]. Acta energiae solaris sinica, 2018, 39(5): 1359-1366.
[4] 林波荣, 李紫微. 面向设计初期的建筑节能优化方法[J]. 科学通报, 2016, 61(1): 113-121.
LIN B R, LI Z W.Building energy-saving approach in early design stage[J]. Chinese science bulletin, 2016, 61(1): 113-121.
[5] 张浩田, 温蜜, 李晋国, 等. 数据驱动的时间注意力卷积风电功率预测模型[J]. 太阳能学报, 2022, 43(10): 167-176.
ZHANG H T, WEN M, LI J G, et al.Data driven time attention convolution wind power prediction model[J]. Acta energiae solaris sinica, 2022, 43(10): 167-176.
[6] 韩超, 宋苏, 王成红. 基于ARIMA模型的短时交通流实时自适应预测[J]. 系统仿真学报, 2004(7): 1530-1532, 1535.
HAN C, SONG S, WANG C H.A real-time short-term traffic flow adaptive forecasting method based on ARIMA model[J]. Journal of system simulation, 2004(7): 1530-1532, 1535.
[7] 葛娜, 孙连英, 石晓达, 等. Prophet-LSTM组合模型的销售量预测研究[J]. 计算机科学, 2019, 46(S1): 446-451.
GE N, SUN L Y, SHI X D, et al.Research on sales forecast of Prophet-LSTM combination model[J]. Computer science, 2019, 46(S1): 446-451.
[8] 何大四, 张旭. 改进的季节性指数平滑法预测空调负荷分析[J]. 同济大学学报(自然科学版), 2005(12): 1672-1676.
HE D S, ZHANG X.Analysis of air conditioning load prediction by modified seasonal exponential smoothing model[J]. Journal of Tongji University(natural science), 2005(12): 1672-1676.
[9] 薛东, 段立强, 高统彤, 等. 考虑多重特征与不确定性度量的综合能源系统负荷预测研究[J]. 太阳能学报, 2024, 45(7): 379-388.
XUE D, DUAN L Q, GAO T T, et al.Study of integrated energy system load forecasting considering multiple characteristics and uncertainty measures[J]. Acta energiae solaris sinica, 2024, 45(7): 379-388.
[10] NOROUZI P, MAALEJ S, MORA R.Applicability of deep learning algorithms for predicting indoor temperatures: towards the development of digital twin HVAC systems[J]. Buildings, 2023, 13(6): 1542.
[11] MA H J, XU L J, JAVAHERI Z, et al.Reducing the consumption of household systems using hybrid deep learning techniques[J]. Sustainable computing: informatics and systems, 2023, 38: 100874.
[12] JEON B K, KIM E J.LSTM-based model predictive control for optimal temperature set-point planning[J]. Sustainability, 2021, 13(2): 894.
[13] 欧阳静, 杨吕, 尹康, 等. 基于ALIF-LSTM多任务学习的综合能源系统短期负荷预测[J]. 太阳能学报, 2022, 43(9): 499-507.
OUYANG J, YANG L, YIN K, et al.Short-term load forecasting method for integrated energy system based on ALIF-LSTM and multi-task learning[J]. Acta energiae solaris sinica, 2022, 43(9): 499-507.
[14] 王曦, 甘灵丽, 王亮, 等. 基于LM-BP神经网络的办公建筑逐时空调能耗预测[J]. 建筑节能(中英文), 2024, 52(1): 58-64.
WANG X, GAN L L, WANG L, et al.Hourly energy consumption prediction of office building air-conditioning based on LM-BP neural network[J]. Journal of BEE, 2024, 52(1): 58-64.
[15] 林志洁, 罗壮, 赵磊, 等. 特征金字塔多尺度全卷积目标检测算法[J]. 浙江大学学报(工学版), 2019, 53(3): 533-540.
LIN Z H, LUO Z, ZHAO L, et al.Multi-scale convolution target detection algorithm with feature pyramid[J]. Journal of Zhejiang University(engineering science), 2019, 53(3): 533-540.
[16] 刘贵喜, 杨万海. 基于多尺度对比度塔的图像融合方法及性能评价[J]. 光学学报, 2001(11): 1336-1342.
LIU G X, YANG W H.A multiscale contrast-pyramid-based image fusion scheme and its performance evaluation[J]. Acta optica sinica, 2001(11): 1336-1342.