Prediction of Residential Building Energy Efficiency Performance using Deep Neural Network

One of the important discussions currently in building energy use is the prediction of energy consumption. To achieve energy savings and reduce environmental impact, the prediction of energy consumption in buildings is crucial to improve energy performance. In this paper, an improved prediction of e...

Full description

Saved in:
Bibliographic Details
Published inIAENG international journal of computer science Vol. 48; no. 3; p. 731
Main Authors Irfan, Muhammad, Ramlie, Faizir, Widianto, Lestandy, Merinda, Faruq, Amrul
Format Journal Article
LanguageEnglish
Published Hong Kong International Association of Engineers 27.08.2021
Subjects
Algorithms
Artificial neural networks
Back propagation networks
Cooling loads
Deep learning
Energy consumption
Energy efficiency
Environmental impact
Heating load
Learning theory
Machine learning
Methods
Multilayer perceptrons
Neural networks
Residential buildings
Residential energy
Root-mean-square errors
Online AccessGet full text

Cover

More Information
Summary:One of the important discussions currently in building energy use is the prediction of energy consumption. To achieve energy savings and reduce environmental impact, the prediction of energy consumption in buildings is crucial to improve energy performance. In this paper, an improved prediction of energy efficiency performance for the heating load (HL) and cooling load (CL) of residential buildings is demonstrated. A deep learning method using a deep neural network (DNN) based on a multilayer feed-forward artificial neural network (ANN) trained with stochastic gradient descent using back-propagation was examined. The proposed DNN method was also compared with a simple multilayer perceptron (MLP) ANN method. The error performances of both DNN and ANN methods were also analyzed against various machine learning algorithms used in previous studies. The results showed that the proposed DNN method performed better in terms of error performance for the mean absolute error (MAE), root mean square error (RMSE), and mean absolute percentage error (MAPE) values compared with the other methods. Adequate values of coefficient of determination (R2) were also obtained for both HL and CL predictions of the proposed DNN method, an indication of good prediction performance. Overall, the proposed ANN and DNN methods proved to outperform the other methods reviewed in this study. Based on these findings, it was concluded that the proposed DNN method was statistically a significant approach within the related research area.
ISSN:1819-656X
1819-9224