制冷技术2025，Vol.45Issue(1)：74-80,7.DOI:10.3969/j.issn.2095-4468.2025.01.205
建筑室内温度时空全域预测方法研究
Research on Spatiotemporal Global Prediction Method for Indoor Temperature in Buildings
史万涛 1李元阳 1方兴 1罗章维 1闫锐1
作者信息
- 1. 广东美的暖通设备有限公司,广东 佛山 528311
- 折叠
摘要
Abstract
In order to achieve rapid and accurate indoor temperature prediction,a temporal and spatial prediction method for indoor temperature is proposed,including data screening,optimal sensor combination and spatial layout,time series prediction and spatial distribution prediction.Among them,box plots are used to screen abnormal data,bottleneck neural networks are used to select the optimal sensor combination and layout,temporal fusion transformers are used to predict the future temperature values of each sensor,and Kriging interpolation is used to predict the spatial temperature distribution.Then the proposed method is validated on a floor of a commercial building in Guangdong province.Results show that the root mean square error of temporal prediction is 0.29,and the spatial prediction error is within 5%,both of which have high prediction accuracy and can be used as a prerequisite for thermal comfort control.关键词
数据驱动/时间和空间预测/神经网络/克里金插值Key words
Data driven/Temporal and spatial prediction/Neural network/Kriging interpolation分类
通用工业技术引用本文复制引用
史万涛,李元阳,方兴,罗章维,闫锐..建筑室内温度时空全域预测方法研究[J].制冷技术,2025,45(1):74-80,7.
