基于GRU神经网络的建筑能耗预测方法研究  

Research onbuilding energy consumption prediction method based on GRU neural network

作　　者：岑健 王付宇 伍银波 刘永桂[2] 苏宇贵 CEN Jian;WANG Fuyu;WU Yinbo;LIU Yonggui;SU Yugui(School cf Automation,Guangdong Polytechnic Normal University,Guangzhou Guangdong 510665,China;School of Automation.Science and Engineering,South China University of Technology,Guangzhou Guangdong 510635,China;Guangzhou Zhongyu Air Conditioning Technology Development Company Limited,Guangzhou Guangdong 510935,China)

机构地区：[1]广东技术师范大学自动化学院,广州510665 [2]华南理工大学自动化科学与工程学院,广州510635 [3]广州中宇冷气科技发展有限公司,广州510935

出　　处：《自动化与仪器仪表》2021年第4期156-159,共4页Automation & Instrumentation

基　　金：广东省应用型研发专项资金项目(No.2016B020243011);广东省普通高校专项项目(No.2020KCXTD017,No.2019KZDZX1004);广州市重点实验室建设项目(No.202002010003)。

摘　　要：针对现有建筑能耗预测方法时序性差导致预测精度低的问题,提出了一种基于门控循环单元(GRU)神经网络的建筑能耗预测方法。首先,对数据进行预处理,使用样本平均值对样本缺失值进行填充,并对数据作标准化操作;然后,基于Keras深度学习框架搭建GRU神经网络,输入数据进行训练和测试,同时使用自适应矩估计(Adam)算法更新和优化模型的权重参数;最后,将GRU神经网络方法与循环神经网络(RNN)方法进行比较,选用平均绝对误差(MAE)、均方根误差(RMSE)和平均绝对百分比误差(MAPE)作为评价指标,实验结果表明,本方法的误差更小,具有更高的预测精度。In order to solve the problem of poor temporality resulting in low prediction accuracy of existing building energy consumption prediction methods,a building energy consumption prediction method based on gated recurrent unit(GRU)neural network is proposed.First of all,the data is preprocessed.The missing data of the samples is filled with the sample mean,and the data is standardized.Then,the GRU neural network is built based on Keras deep learning framework,and the input data is used for training and testing.At the same time,adaptive moment estimation(Adam)algorithm is used to update and optimize the weight parameters of the model.Finally,the GRU neural network method is compared with recurrent neural network(RNN)method,and mean absolute error(MAE),root mean square error(RMSE)and mean absolute percentage error(MAPE)are selected as indicators.The experimental results show that the method of this paper has smaller error and higher precision of prediction.

关 键 词：建筑能耗预测 时间序列 深度学习 GRU神经网络 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]