制冷技术2024，Vol.44Issue(6)：50-58,74,10.DOI:10.3969/j.issn.2095-4468.2024.06.202
基于生成对抗网络学习的建筑暖通空调负荷特征能耗研究
Study on Energy Consumption of Building HVAC Load Characteristics Based on Generative Adversarial Network Learning
摘要
Abstract
The generative adversarial networks(GAN)and back-propagation neural network(BPNN)are adopted to explore the problem of low energy consumption data in the initial stage of building construction or renovation.The performance of a prediction model based on a mixed dataset and a prediction model based on raw data is compared by constructing an energy consumption prediction model.The results show that compared with the energy consumption data collected in heating,ventilation,and air conditioning(HVAC)system,the energy consumption data generated by the GAN successfully learns the latter's energy consumption load characteristics,generates samples that are not in the original dataset.The average absolute error of the prediction model constructed from the mixed dataset on the test dataset is 4.086×10-2,which is a maximum reduction of 11.10%compared with the prediction model constructed from the original dataset,which indicates that the prediction model after data expansion is more effective,validates the feasibility of GAN for energy consumption prediction data augmentation.关键词
机器学习/能耗特征学习/生成对抗网络/数据增强/空调系统能耗预测Key words
Machine learning/Energy profile synthesis/Generative adversarial Networks/Data argument/Energy consumption prediction分类
通用工业技术引用本文复制引用
许源驿,赵思哲,陈焕新,王琴,程亨达..基于生成对抗网络学习的建筑暖通空调负荷特征能耗研究[J].制冷技术,2024,44(6):50-58,74,10.基金项目
国家自然科学基金(No.51876070),大学生创新创业项目(No.NY2021091). （No.51876070）
