A soft computing method for the prediction of energy performance of residential buildings. (October 2017)

• Record Type: Journal Article
• Authors: Nilashi, Mehrbakhsh; Dalvi-Esfahani, Mohammad; Ibrahim, Othman; Bagherifard, Karamollah; Mardani, Abbas; Zakuan, Norhayati
• Is Part Of: Measurement. Volume 109(2017)
• Page Start: 268
• Page End: 280
• Publication Date: 2017-10
• Subjects: Residential buildings -- Adaptive-Network-based Fuzzy Inference System -- PCA -- Estimation of energy efficiency -- Heating load -- Cooling load; Weights and measures -- Periodicals; Measurement -- Periodicals; Measurement; Weights and measures; Periodicals; 530.8
• DOI: 10.1016/j.measurement.2017.05.048
• Languages: English
• ISSNs: 0263-2241

Abstract

Graphical abstract: Highlights: A method is proposed for energy performance prediction of residential buildings. The method is developed using EM, PCA and ANFIS. Energy efficiency dataset obtained from UCI is used in evaluating the method. The MAE of the predictions for HL and CL are respectively 0.16 and 0.52. Abstract: Buildings are a crucial factor of energy concerns and one of the most significant energy consumers. Accurate estimation of energy efficiency of residential buildings based on the computation of Heating Load (HL) and the Cooling Load (CL) is an important task. Developing computational tools and methods for prediction of energy performance will help the policy makers in efficient design of building. The aim of this study is therefore to develop an efficient method for the prediction of energy performance of residential buildings using machine learning techniques. Our method is developed through clustering, noise removal and prediction techniques. Accordingly, we use Expectation Maximization (EM), Principal Component Analysis (PCA) and Adaptive Neuro-Fuzzy Inference System (ANFIS) methods for clustering, noise removal and prediction tasks, respectively. Experimental results on real-world dataset show that proposed method remarkably improves the accuracy of prediction in relation to the existing state-of-the-art techniques and is efficient in estimating the energy efficiency of residential buildings. The Mean Absolute Error (MAE) of the predictions for HL and CL are respectively 0.16 and 0.52 which show the effectiveness of our method in predicting HL and CL.