作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2019, Vol. 45 ›› Issue (6): 290-296. doi: 10.19678/j.issn.1000-3428.0050649

• 开发研究与工程应用 • 上一篇    下一篇

基于AR-DBN的建筑分项能耗短期预测

钱青1,唐桂忠1,2,张广明1,2,邓歆1,尹海培3   

  1. 1.南京工业大学 电气工程与控制科学学院,南京 211800; 2.江苏省绿色建筑工程技术研究中心,南京 211800;3.江苏省住房和城乡建设厅 科技发展中心,南京 211800
  • 收稿日期:2018-03-07 出版日期:2019-06-15 发布日期:2019-06-15
  • 作者简介:钱青(1989—),女,硕士研究生,主研方向为建筑能耗预测与控制;唐桂忠,副教授、博士;张广明,教授、博士;邓歆,讲师、博士;尹海培,工程师、硕士。
  • 基金资助:
    国家自然科学基金(51507078);江苏省教育厅高校自然科学基金(15KJB470006)。

Short term prediction of itemized building energy consumption based on AR-DBN

QIAN Qing1,TANG Guizhong1,2,ZHANG Guangsming1,2,DENG Xin1,YIN Haipei3   

  1. 1.School of Electrical Engineering and Control Science,Nanjing Tech University,Nanjing 211800,China;2.Jiangsu Green Building Engineering Technology Research Center,Nanjing 211800,China;3.Science and Technology Development Center,Jiangsu Housing and Urban-Rural Development Department,Nanjing 211800,China
  • Received:2018-03-07 Online:2019-06-15 Published:2019-06-15

摘要: 针对现有总能耗预测方法不能准确分辨建筑能耗的消耗去向且预测精度较低的问题,根据能耗用途,将总能耗分为4项,提出一种建筑能耗分项预测模型。基于时间序列自回归模型,对建筑物的照明能耗进行短期预测。构建深度置信网络模型,根据照明能耗预测结果、室外逐时平均温度、室外逐时平均相对湿度、天气特征值、节假日、逐时平均风速以及一天24个整点时刻,分项预测空调能耗、动力能耗和特殊能耗。实验结果表明,相比总能耗预测模型iPSO-BP和BP,该模型能更加精确、有效地预测建筑能耗中的各分项能耗。

关键词: 时间序列, 自回归模型, 分项能耗, 深度学习, 深度置信网络

Abstract: To address the problem that the existing methods of total energy consumption prediction cannot accurately distinguish where the consumption of building energy is consumed and have the low prediction.To this end,according to the use of energy consumption,this paper divides the total energy consumption into four items and proposes a prediction model of itemized building energy consumption.Firstly,based on time sequence Auto-regression(AR) model,the lighting energy consumption of a building is predicted in short term.Secondly,a Deep Belief Network(DBN) model is constructed,which predicts air conditioning energy consumption,power energy consumption and special energy consumption of the building according to the lighting energy consumption prediction results,hourly average outdoor temperature,hourly average outdoor relative humidity,weather feature,holiday,hourly average wind speed,time.Experimental results show that,compared with the total energy consumption prediction models iPSO-BP and BP,the proposed model can predict the itemized energy consumption of a building more accurately and effectively.

Key words: time sequence, Auto-regression(AR) model, itemized energy consumption, deep learning, Deep Belief Network(DBN)

中图分类号: