基于长短期记忆神经网络的冷水机组能耗预测  

Energy Consumption Prediction of Chiller Based on Long Short-Term Memory

作　　者：徐麟 周传辉[1] 胡云鹏[2] 李冠男 方曦 Xu Lin;Zhou Chuanhui;Hu Yunpeng;Li Guannan;Fang Xi(Wuhan University of science and technology,Wuhan,430081;Wuhan business school,Wuhan,430056)

机构地区：[1]武汉科技大学,武汉430081 [2]武汉商学院,武汉430056

出　　处：《制冷与空调（四川）》2020年第6期664-669,共6页Refrigeration and Air Conditioning

基　　金：国家自然科学基金资助项目(编号:51906181)。

摘　　要：空调系统中冷水机组是主要的耗能部件,节能潜力较大。由于空调系统的末端需求的变化性,准确开展冷水机组的能耗预测能有效的为机组的优化控制提供参考。因此,引入长短期记忆神经网络(LSTM,Long Short-Term Memory)对冷水机组能耗进行预测,并结合EnergyPlus仿真模型数据和实际办公建筑数据验证LSTM能耗预测模型的预测效果。建立并优化LSTM冷水机组能耗预测模型。结果显示,相比于反向传播神经网络和多元线性回归模型,LSTM模型的计算时间有所增加,但LSTM模型的预测精度在三个模型中最高,LSTM能够更准确的预测冷水机组的能耗。The chiller is the main energy consuming part in the air conditioning system,which has great energy saving potential.Due to the change of terminal demand of air conditioning system,accurate prediction of energy consumption of water chiller can effectively provide reference for optimal control of the unit.Therefore,this paper introduces long short term memory(LSTM)to predict the energy consumption of water chillers,and validates the prediction effect of LSTM combined with the data of EnergyPlus simulation model and actual office building data.After the data is divided into training set and test set,the data is standardized.The energy consumption prediction model of LSTM chiller is established and optimized.The results show that compared with the back propagation neural network model and the multiple linear regression model,the calculation time of the LSTM model is increased,but the prediction accuracy of the LSTM model is the highest among the three models,and the LSTM can predict the energy consumption of the chiller more accurately.

关 键 词：冷水机组 能耗预测 长短期记忆神经网络 时间序列 

分 类 号：TU83[建筑科学—供热、供燃气、通风及空调工程]