摘要
为了提高建筑能耗预测精度和效率,以某城市商业建筑为研究对象,收集建筑能耗基础数据,从缺失值、异常值、归一化等方面进行数据预处理。用Adam优化器对构建的深度学习模型进行训练,并基于均方误差(mean square error,MSE)对深度学习模型的预测结果进行验证。验证结果表明,迭代次数取60次,时间序列窗口取6时,长短期记忆(long short-term memory,LSTM)模型和事件驱动架构-长短期记忆(event-driven architecture-long short-term memory,EDA-LSTM)模型的预测精度和效率最佳,且EDA-LSTM模型对建筑能耗的预测更准确。
In order to improve the accuracy and efficiency of building energy consumption prediction,this paper takes a commercial building in a city as the research object.collects the basic data of building energy consumption,and preprocesses the data from the aspects of missing value,outlier and normalization.The Adam optimizer is used to train the constructed deep learning model,and the prediction results from the deep learning model are verified based on the mean square error(MSE).The verification results show that when the number of iterations is set to 60 and the timeseries window is set to 6,the long short-term memory(LSTM)model and the event-driven architecture-long short-term memory(EDA-LSTM)model achieve their optimal predictive accuracy and efficiency,and the EDALSTM model demonstrates greater accuracy in predicting building energy consumption.
作者
张益弛
ZHANG Yichi(Meishan Vocational&Technical College,Meishan 620000,Sichuan,China)
出处
《工程技术研究》
2025年第24期214-216,共3页
Engineering and Technological Research
基金
四川省眉山职业技术学院校级科研课题“基于LSTM长短期网络深度学习方法的建筑能耗预测研究”(23KY08)。
