西安建筑科技大学学报(自然科学版)

图书馆在高校建筑中具有非常重要的地位,有较大的节能潜力.然而,近年来对于高校图书馆建筑节能的研究偏少,本文通过提出一种基于相似日LM(Levenberg-Marquardt)神经网络的高校图书馆能耗预测模型,为高校图书馆能耗研究提供参考.以我国某高校图书馆为例,首先通过统计分析的方法得到影响图书馆能耗较为重要的因素,即室内人员、开放策略及气温.然后利用模糊聚类法划分相似日,依据相似日将原有数据进行筛选.接着将处理后的数据对预测模型进行训练.最后将改进后的预测模型与原预测模型的各项指标进行对比分析.依据对比结果可知,改进后模型的平均绝对百分比误差和均方误差分别降低了1.28%和23.06,拟合度提高了0.0421.

As a major part of campus buildings, libraries have great potential of energy-saving. But there are very few recent studies focused on energy conservation in university library buildings. This paper explores energy use patterns in library buildings and proposes a energy consumption prediction model for campus libraries based on Levenberg-Marquardt neural network. The model may provide reference to future energy-saving efforts. A university library building in China was chosen as the research subject. By analyzing related data, we are able to find the factors affecting energy consumption of the library most, which are occupancy, outdoor air temperature, as well as the opening hours of the library. The similar day method was employed to train the model: Through soft clustering, the energy consumption profiles of each day was put into categories which were in turn used to as inputs. Compared with the traditional LM model, the optimized LM neural network had showed a decrease in absolute percentage error and mean square error by 1.28%and 23.06, respectively, and the fitting degree had been improved by 0.042 1.

引言
1 预测算法原理
2 实例分析
3 模型的建立
4 结果验证
5 结语

表1 暑假图书馆开放时间表<br/>Tab.1 Library opening schedule during summer vacation

表1 暑假图书馆开放时间表
Tab.1 Library opening schedule during summer vacation

表2 学期图书馆开放时间表<br/>Tab.2 Library opening schedule during terms

表2 学期图书馆开放时间表
Tab.2 Library opening schedule during terms

图1 图书馆室内人数与能耗<br/>Fig.1 Occupancy and energy consumption

图1 图书馆室内人数与能耗
Fig.1 Occupancy and energy consumption

图2 图书馆室外日均气温<br/>Fig.2 Average daily outdoor temperature

图2 图书馆室外日均气温
Fig.2 Average daily outdoor temperature

表3 图书馆室内人员与能耗的R2统计表<br/>Tab.3 R-square value of the occupancy and the energy consumption in the library

表3 图书馆室内人员与能耗的R2统计表
Tab.3 R-square value of the occupancy and the energy consumption in the library

图3 预测流程
Fig.3 Prediction process

表4 图书馆开放区域量化表<br/>Tab.4 Quantification table of library open areas

表4 图书馆开放区域量化表
Tab.4 Quantification table of library open areas

表5 图书馆开放时间量化<br/>Tab.5 Quantification table of library open hours

表5 图书馆开放时间量化
Tab.5 Quantification table of library open hours

图4 各隐含层层数迭代次数和均方误差<br/>Fig.4 The iterations and the MSE value of different hidden layers

图4 各隐含层层数迭代次数和均方误差
Fig.4 The iterations and the MSE value of different hidden layers

图5 各隐含层节点数迭代次数和均方误差<br/>Fig.5 The iterations and the MSE value of different hidden layer nodes

图5 各隐含层节点数迭代次数和均方误差
Fig.5 The iterations and the MSE value of different hidden layer nodes

表6 模型预测结果评价指标对比<br/>Tab.6 Comparison of the two models' prediction indicators

表6 模型预测结果评价指标对比
Tab.6 Comparison of the two models' prediction indicators

图6 相似日LM神经网络模型拟合情况<br/>Fig.6 The comparison of the fitting value of the optimized LM model and the true value

图6 相似日LM神经网络模型拟合情况
Fig.6 The comparison of the fitting value of the optimized LM model and the true value

图7 普通LM神经网络模型拟合情况<br/>Fig.7 The comparison of the fitting value of the general LM model and the true value

图7 普通LM神经网络模型拟合情况
Fig.7 The comparison of the fitting value of the general LM model and the true value

[1]谭洪卫,徐钰琳,胡承益,等.全球气候变化应对与我国高校校园建筑节能监管[J].建筑热能通风空调,2010,29(1):36-40.
[2]YAN Ding, WANG Qiaochu, WANG Zhaoxia, et al. An occupancy-based model for building electricity consumption prediction: A case study of three campus buildings in Tianjin[J]. Energy & Buildings,2019,202:109412.
[3]高彪,谭洪卫,宋亚超.高校校园建筑用能现状及存在问题分析——以长三角地区某综合型大学为例[J].建筑节能,2011,39(2):41-44.
[4]CHUNG Min Hee,RHEE Eon Ku. Potential opportunities for energy conservation in existing buildings on university campus: A field survey in Korea[J]. Energy & Buildings,2014,78: 176-182.
[5]WALTER Leal Filho, AMANDA Lange Salvia, ARMINDA do Paço, et al. A comparative study of approaches towards energy efficiency and renewable energy use at higher education institutions[J]. Journal of Cleaner Production,2019,237:117728.
[6]WANG Y, SHAO L. Understanding occupancy pattern and improving building energy efficiency through Wi-Fi based indoor positioning[J].Building and Environment, 2017, 45(3): 53.
[7]王茜. 西北某高校图书馆能耗预测模型与节能策略研究[D].西安:西安建筑科技大学,2021.
[8]YIK F W H,Burnett J,Prescott I. Predicting air-conditioning energy consumption of a group of buildings using different heat rejection methods[J]. Energy & Buildings,2001,33(2): 151-166
[9]SEPEHR M, EGHTEDAEI R, TOOLABIMOGHADAM A, et al. Modeling the electrical energy consumption profile for residential buildings in Iran[J]. Sustainable Cities and Society, 2018, 41(10): 481-489.
[10]MERIH Aydinalp,V. ISMET Ugursal,ALAN S. Fung. Modeling of the appliance, lighting, and space-cooling energy consumptions in the residential sector using neural networks[J]. Applied Energy,2002,71(2): 87-110
[11]ALBERTO Hernandez Neto, FLÁVIO Augusto Sanzovo Fiorelli. Comparison between detailed model simulation and artificial neural network for forecasting building energy consumption[J]. Energy & Buildings,2008,40(12): 2169-2176
[12]LI Q, REN P, MENG Q. Prediction model of annual energy consumption of residential buildings[C]//2010 International Conference on Advances in Energy Engineering. Beijing:ICAEE,2010.
[13]谈力. 基于相似日选取的小波极限学习机短期负荷预测模型研究[D].南京:南京理工大学,2015.
[14]张悦. 某大型公共建筑能耗分形特性与预测模型研究[D].西安:西安建筑科技大学,2019.
[15]黄豪彩,黄宜坚,杨冠鲁. 基于LM算法的神经网络系统辨识[J]. 组合机床与自动化加工技术, 2003(2): 6-8.
[16]QIAO J,WANG L,YANG C,et al. Adaptive leven bergmarquardt algorithm based echo state network for chaotic time series prediction[ J]. IEEE Access,2018,6:10720-10732.
[17]杨雪. 基于深度学习的大型公建能耗预测及信息管理系统研究[D].西安:西安建筑科技大学,2019.

备注

引言

1 预测算法原理

2 实例分析

3 模型的建立

4 结果验证

5 结语

期刊信息