Modelling the energy performance of residential buildings using advanced computational frameworks based on RVM, GMDH, ANFIS-BBO and ANFIS-IPSO. (March 2021)

• Record Type: Journal Article
• Authors: Kardani, Navid; Bardhan, Abidhan; Kim, Dookie; Samui, Pijush; Zhou, Annan
• Is Part Of: Journal of building engineering. Volume 35(2021)
• Publication Date: 2021-03
• Subjects: Energy performance of buildings -- Hybrid computational models -- Heating load -- Cooling load -- Statistical analysis; Building -- Periodicals; 690.05
• DOI: 10.1016/j.jobe.2020.102105
• Languages: English
• ISSNs: 2352-7102

Abstract

Abstract: Modelling the heating load (HL) and cooling load (CL) is the cornerstone of the designing of energy-efficient buildings, since it determines the heating and cooling equipment requirements needed to retain comfortable indoor air conditions. Advanced and specialised modelling tools for energy-efficient buildings may provide a reliable estimation of the effect of alternative building designs. However, implementing these tools can be a labour-intensive task, very time-consuming and dependent on user experiences. Hence, in this study, four advanced computational frameworks including relevance vector machine (RVM), group method of data handling (GMDH), hybridization of adaptive neuro-fuzzy interface system (ANFIS) and biogeography-based optimisation (BBO), i.e. ANFIS-BBO, and hybridization of ANFIS and improved particle swarm optimisation (IPSO), i.e. ANFIS-IPSO, are proposed as novel approaches to predict the heating load (HL) and cooling load (CL) of residential buildings. Obtained results from the proposed models are compared using several performance parameters. In addition, several visualisation methods including Taylor diagram, regression characteristic curve, a novel method called accuracy matrix and rank analysis are used to demonstrate the model with the best performance. Furthermore, Anderson–Darling' Normality (A-D) test and Mann–Whitney U' (M − W) tests are studied as non-parametric statistical test for further investigations of the models. Obtained results indicate the excellent ability of the applied models to map the non-linear relationships between the input and output variables. Result also identified RVM as the best predictive model among four proposed models. Finally, two equations are derived from the RVM model to address the HL and CL of residential buildings. Highlights: Energy performance of buildings (EPB) is modelled based on prediction of heating and cooling loads. Four advanced computational frameworks are developed based on RVM, GMDH, ANFIS-BBO and ANFIS-IPSO. Various statistical analysis are employed to evaluate the obtained results from the models. Obtained results indicate the excellent ability of the proposed models. RVM outperforms other advanced computational models in predicting both HL and CL.