基于CNN-RNN组合模型的办公建筑能耗预测

doi:10.16183/j.cnki.jsjtu.2021.192

上海交通大学学报 ›› 2022, Vol. 56 ›› Issue (9): 1256-1261.doi: 10.16183/j.cnki.jsjtu.2021.192

所属专题：《上海交通大学学报》2022年“机械与动力工程”专题

基于CNN-RNN组合模型的办公建筑能耗预测

曾国治¹, 魏子清¹, 岳宝², 丁云霄², 郑春元², 翟晓强¹()

1.上海交通大学机械与动力工程学院,上海 200240
2.广东美的暖通设备有限公司,广东佛山 528311

收稿日期:2021-05-23 出版日期:2022-09-28 发布日期:2022-10-09
通讯作者: 翟晓强 E-mail:xqzhai@sjtu.edu.cn
作者简介:曾国治(1998-),男,四川省泸州市人,硕士生,主要从事智慧建筑与建筑节能研究.

Energy Consumption Prediction of Office Buildings Based on CNN-RNN Combined Model

ZENG Guozhi¹, WEI Ziqing¹, YUE Bao², DING Yunxiao², ZHENG Chunyuan², ZHAI Xiaoqiang¹()

1. School of Mechanical Engineering, Shanghai Jiao Tong University, Shanghai 200240, China
2. Guangdong Midea HVAC Equipment Co., Ltd., Foshan 528311, Guangdong, China

Received:2021-05-23 Online:2022-09-28 Published:2022-10-09
Contact: ZHAI Xiaoqiang E-mail:xqzhai@sjtu.edu.cn

摘要/Abstract

摘要：

为准确反映办公建筑的运行特性,利用卷积神经网络(CNN)良好的特征提取能力与循环神经网络(RNN)良好的时序学习能力,提出用于预测办公建筑能耗的CNN-RNN组合模型,并对应设计了适用于深度学习模型的二维矩阵数据输入结构.案例分析结果表明,相较于简单循环神经网络和长短期记忆网络,CNN-RNN组合模型的预测精度与计算效率均显著提升,模型泛化性好.

关键词: 建筑能耗预测, 卷积神经网络, 循环神经网络, 深度学习

Abstract:

In order to accurately reflect the operation characteristics of office buildings, a convolutional neural network(CNN)-recurrent neural network(RNN)combined model for energy consumption prediction of office buildings is proposed by using the good feature extraction ability of CNN and the good time series learning ability of RNN. Besides, a two-dimensional matrix data input structure suitable for the deep learning model is designed. The case study results show that compared with the simple recurrent neural network and long short term memory network, both the prediction accuracy and computational efficiency of CNN-RNN combined model are significantly improved, and the generalization of the model is also good.

Key words: prediction of building energy consumption, convolutional neural network (CNN), recurrent neural network(RNN), deep learning

中图分类号:

TU855

曾国治, 魏子清, 岳宝, 丁云霄, 郑春元, 翟晓强. 基于CNN-RNN组合模型的办公建筑能耗预测[J]. 上海交通大学学报, 2022, 56(9): 1256-1261.

ZENG Guozhi, WEI Ziqing, YUE Bao, DING Yunxiao, ZHENG Chunyuan, ZHAI Xiaoqiang. Energy Consumption Prediction of Office Buildings Based on CNN-RNN Combined Model[J]. Journal of Shanghai Jiao Tong University, 2022, 56(9): 1256-1261.

图/表 10

图1

图2

表1

图3

表2

表3

表4

图4

图5

图6

参考文献 14

[1]	中国建筑节能协会. 中国建筑能耗研究报告2020[J]. 建筑节能, 2021, 49(2): 1-6.
	China Association of Building Energy Efficiency. China building energy consumption annual report 2020[J]. Building Energy Efficiency, 2021, 49(2): 1-6.
[2]	FOUCQUIER A, ROBERT S, SUARD F, et al. State of the art in building modelling and energy performances prediction: A review[J]. Renewable and Sustainable Energy Reviews, 2013, 23: 272-288. doi: 10.1016/j.rser.2013.03.004 URL
[3]	GASSAR A A A, CHA S H. Energy prediction techniques for large-scale buildings towards a sustainable built environment: A review[J]. Energy and Buildings, 2020, 224: 110238.
[4]	FU X, ZENG X J, FENG P P, et al. Clustering-based short-term load forecasting for residential electricity under the increasing-block pricing tariffs in China[J]. Energy, 2018, 165: 76-89. doi: 10.1016/j.energy.2018.09.156 URL
[5]	KIM J H, SEONG N C, CHOI W. Forecasting the energy consumption of an actual air handling unit and absorption chiller using ANN models[J]. Energies, 2020, 13(17): 4361.
[6]	ROBINSON C, DILKINA B, HUBBS J, et al. Machine learning approaches for estimating commercial building energy consumption[J]. Applied Energy, 2017, 208: 889-904. doi: 10.1016/j.apenergy.2017.09.060 URL
[7]	RAJAGUKGUK R A, RAMADHAN R A A, LEE H J. A review on deep learning models for forecasting time series data of solar irradiance and photovoltaic power[J]. Energies, 2020, 13(24): 6623. doi: 10.3390/en13246623 URL
[8]	FAN C, SUN Y J, ZHAO Y, et al. Deep learning-based feature engineering methods for improved building energy prediction[J]. Applied Energy, 2019, 240: 35-45. doi: 10.1016/j.apenergy.2019.02.052 URL
[9]	廖文强, 王江宇, 陈焕新, 等. 基于深度学习的建筑能耗短期预测方法研究[C]// 2019年中国家用电器技术大会论文集. 佛山: 电器杂志社, 2019: 1173-1178
	LIAO Wenqiang, WANG Jiangyu, Chen Huanxin, et al. Research on short-term building energy consumption prediction method based on deep-learning[C]// Proceeding of 2019 China Household Electrical Appliances Technical Conference. Foshan, China: Appliance, 2019: 1173-1178.
[10]	邹锋, 田大伟, 王悦, 等. 基于生成对抗网络的深度学习能耗预测算法[J]. 电脑知识与技术, 2019, 15(2): 198-200.
	ZOU Feng, TIAN Dawei, WANG Yue, et al. Deep learning energy consumption prediction algorithms based on generative adversarial networks[J]. Computer Knowledge and Technology, 2019, 15(2): 198-200.
[11]	RUNGE J, ZMEUREANU R. A review of deep learning techniques for forecasting energy use in buildings[J]. Energies, 2021, 14(3): 608. doi: 10.3390/en14030608 URL
[12]	YAMASHITA R, NISHIO M, DO R K G, et al. Convolutional neural networks: An overview and application in radiology[J]. Insights into Imaging, 2018, 9(4): 611-629. doi: 10.1007/s13244-018-0639-9 pmid: 29934920
[13]	HAZRA A, CHOUDHARY P, SHEETAL S M. Recent advances in deep learning techniques and its applications: An overview[C]// Advances in Biomedical Engineering and Technology. Singapore: Springer, 2020: 103-122.
[14]	Iowa State University. Iowa environmental mesonet[DB/OL]. (2019-10-22) [2021-03-16]. http://mesonet.agron.iastate.edu/request/download.phtml?network=CN__ASOS#.

网络层序号	类型	神经元个数	激活函数
1	Conv_1D	64	ReLU
2	Conv_1D	64	ReLU
3	SRNN(返回序列)	64	ELU
4	SRNN	32	ELU
5	Dense	32	ReLU
6	Dense	16	ReLU
7	Dense	8	ReLU

时间轴	模型输出能耗数据/ (kW·h)	模型输入
时间轴	模型输出能耗数据/ (kW·h)	室外气温/℃	相对湿度/%	工作日	时刻	历史序列
t+1	W_t₊₁	T_t₊₁	φ_t₊₁	D_t₊₁	H_t₊₁	W_t_-23
t		T_t	φ_t	D_t	H_t	W_t_-24
t-1		T_t_-1	φ_t_-1	D_t_-1	H_t_-1	W_t_-25
t-2		T_t_-2	φ_t_-2	D_t_-2	H_t_-2	W_t_-26
…		…	…	…	…	…
t-45		T_t_-45	φ_t_-45	D_t_-45	H_t_-45	W_t_-69
t-46		T_t_-46	φ_t_-46	D_t_-46	H_t_-46	W_t_-70
t-47		T_t_-47	φ_t_-47	D_t_-47	H_t_-47	W_t_-71

模型	可训练参数	训练时间/s
SRNN	17596	1193
LSTM	65089	1745
CNN-RNN	17633	848

模型	数据集	MAE/(kW·h)	MAPE/%	R²
SRNN	训练集	28.2	15.38	0.956
	测试集	31.3	15.93	0.946
LSTM	训练集	51.8	25.92	0.859
	测试集	56.2	26.54	0842
CNN-RNN	训练集	32.4	12.97	0.934
	测试集	36.1	13.73	0.918

[1]	. 基于改进YOLOv5l的交通信号灯识别[J]. J Shanghai Jiaotong Univ Sci, 2026, 31(2): 319-333.
[2]	郭琦, 闫军, 郝乾鹏, 韩东, 杨志豪, 闫馨月, 张海鹏, 李然. 基于闭环聚类和多目标优化的风电短期功率预测方法[J]. 上海交通大学学报, 2026, 60(2): 246-255.
[3]	陈亮汶, 朱宇昕, 沈涛, 俞羿帆, 凌霄, 盛庆红. 基于红外尾迹匹配的舰船目标检测算法[J]. 空天防御, 2026, 9(1): 80-90.
[4]	罗志军, 王健瑞, 殷佳伟. 复杂战场环境下的任务驱动智能目标识别方法综述[J]. 空天防御, 2026, 9(1): 1-11.
[5]	李湘, 陈思远, 张俊, 柯德平, 高杰迈, 杨欢欢. 基于物理信息嵌入的非固定长度电力系统暂态稳定快速评估[J]. 上海交通大学学报, 2025, 59(7): 962-970.
[6]	荣光, 张业鑫, 唐朝, 陈金宝, 周奕玲, 王建园. 基于仿真数据驱动的无人飞行器故障诊断技术研究[J]. 空天防御, 2025, 8(6): 73-84.
[7]	曾金灿, 何耿生, 李姚旺, 杜尔顺, 张宁, 朱浩骏. 基于卷积神经网络与轻量级梯度提升树组合模型的电力行业短期以电折碳方法[J]. 上海交通大学学报, 2025, 59(6): 746-757.
[8]	. 基于三维卷积特征金字塔网络的高光谱卫星图像分类[J]. J Shanghai Jiaotong Univ Sci, 2025, 30(6): 1073-1084.
[9]	. 基于深度学习序列方法的多人姿态估计用来检测人体与关键点位置[J]. J Shanghai Jiaotong Univ Sci, 2025, 30(6): 1103-1113.
[10]	潘美琪, 贺兴. 基于零样本学习的风力机故障诊断方法[J]. 上海交通大学学报, 2025, 59(5): 561-568.
[11]	. 基于CEEMDAN 和 GRU的停车位预测[J]. J Shanghai Jiaotong Univ Sci, 2025, 30(5): 962-975.
[12]	. 基于ALBERT的中国诗酒文化命名实体识别[J]. J Shanghai Jiaotong Univ Sci, 2025, 30(5): 1065-1072.
[13]	. 面向太阳能电池复杂缺陷检测的新型多步深度学习方法[J]. J Shanghai Jiaotong Univ Sci, 2025, 30(5): 1050-1064.
[14]	夏伊琳, 刘刚, 鄢丛强, 蔡云泽. 基于深度学习的SAR图像舰船尾迹旋转框检测算法研究[J]. 空天防御, 2025, 8(5): 64-74.
[15]	许强, 马跃华, 许可, 潘俊. 雷达目标智能识别方法研究综述[J]. 空天防御, 2025, 8(5): 1-9.

基于CNN-RNN组合模型的办公建筑能耗预测

Energy Consumption Prediction of Office Buildings Based on CNN-RNN Combined Model

RichHTML

PDF (PC)

可视化

被引次数

摘要/Abstract

引用本文

使用本文

图/表 10

参考文献 14

相关文章 15

编辑推荐

Metrics

本文评价