摘要
针对传统BP神经网络模型预测办公建筑能耗时预测精度受输入变量之间非线性特性影响较大、模型泛化能力不强等问题,建立了结合因子分析法的改进FA-BP神经网络组合预测模型。以实际调研的能耗数据为基础,采用因子分析法对能耗的相关影响因素进行降维处理,得到4个公因子,以此作为神经网络的输入参数进行实证分析。结果显示,改进的FA-BP神经网络组合模型在预测办公建筑能耗方面,较传统的BP神经网络模型具有更高的预测精度,其平均相对误差为1.728%,表明因子分析法可以有效降低神经网络输入参数的复杂维度。
Aiming at the problems of traditional BP neural network model for predicting office building energy consumption that the prediction accuracy is greatly affected by the non-linear characteristics between input variables and the generalization ability of the model is not strong, establishes an improved FA-BP neural network combination prediction model combined with the factor analysis method. Based on the actual energy consumption data, uses the factor analysis method to reduce the dimensions of the related influence factors of energy consumption, and obtains four common factors, which are used as input parameters of the neural network for empirical analysis. The results show that the improved FA-BP neural network combination model has a higher prediction accuracy than the traditional BP neural network model in the prediction of office building energy consumption, and its average relative error is 1.728%, which shows that the factor analysis method can effectively reduce the complex dimension of the input parameters of neural network.
作者
张露
李永安
王德晔
楚广明
刘学来
Zhang Lu;Li Yongan;Wang Deye;Chu Guangming;Liu Xuelai(Shandong Jianzhu University,Jinan,China;不详)
出处
《暖通空调》
2021年第10期125-130,共6页
Heating Ventilating & Air Conditioning
基金
山东省墙材革新与建筑节能重大专项(编号:2014QG012)。
关键词
办公建筑
能耗预测
因子分析
公因子
FA-BP神经网络
组合模型
office building
energy consumption prediction
factor analysis
common factor
FA-BP neural network
combination model