Energy and Buildings 138 (2017) 240–256 Contents lists available at ScienceDirect Energy and Buildings j ourna l ho me pa g e: www.elsevier.com/locate/enbuild A relevant data selection method for energy consumption prediction of low energy building based on support vector machine Subodh Paudel a,b,c , Mohamed Elmitri b , Stéphane Couturier b , Phuong H. Nguyen c , René Kamphuis c,1 , Bruno Lacarrière a , Olivier Le Corre a,∗ a Department of Energy Systems and Environment, Ecole des Mines de Nantes, GEPEA, CNRS, UMR 6144, Nantes, France b Veolia Recherche et Innovation (VERI), Limay, France c Department of Electrical Engineering, Eindhoven University of Technology, Eindhoven, The Netherlands a r t i c l e i n f o Article history: Received 24 March 2016 Received in revised form 29 September 2016 Accepted 8 November 2016 Available online 21 December 2016 Keywords: Building energy consumption Prediction Low energy building Support vector machine Online and offline learning a b s t r a c t Low energy buildings (LEBs) are being considered as a promising solution for the built environment to satisfy high-energy efficiency standards. The technology is based on lowering the overall heat transmis- sion coefficient value (U-value) of the buildings envelope and increasing a heat capacity thus creating a higher thermal inertia. However, LEB introduces a large time constant compared to conventional building due to which it slows the rate of heat transfer between interior of building and outdoor environment and alters the indoor climate regardless of sudden changes in climatic conditions. Therefore, it is challenging to estimate and predict thermal energy demand for such LEBs. This work focuses on artificial intelligence (AI) model to predict energy consumption of LEB. Two kinds of AI modeling approaches: “all data” and “relevant data” are considered. The “all data” uses all available training data and “relevant data” uses a small representative day dataset and addresses the complexity of building non-linear dynamics by introducing past day climatic impacts behavior. This extraction is based on dynamic time warping pattern recognition methods. The case study consists of a French residential LEB. The numerical results showed that “relevant data” modeling approach that relies on small representative data selection has higher accuracy (R 2 = 0.98; RMSE = 3.4) than “all data” modeling approach (R 2 = 0.93; RMSE = 7.1) to predict heating energy load. © 2016 Elsevier B.V. All rights reserved. 1. Introduction The energy efficiency of buildings has drawn significant atten- tion in the recent years as a tool to reach a reduction of energy consumption. According to the European Union (EU), the building sector represents 40% of the total energy consumption resulting in 36% of CO 2 gas emissions [1]. It is estimated that residential build- ings account for 25% of the final energy consumption in the EU [1]. This energy consumption varies depending on the materials used in the walls and roofs of buildings affecting the heat transfer mech- anisms. Low heat transfer materials in the building’s envelope help to improve the energy efficiency resulting in low energy building ∗ Corresponding author. E-mail addresses: Olivier.Lecorre@mines-nantes.fr, lecorre@mines-nantes.fr, ollecorre@gmail.com (O. Le Corre). 1 R. Kamphuis works at TUe, a part of VSNU. (LEB) or passive house, by lowering the heat transfer coefficient (U-value). In the case of a LEB, the high insulation level increases the impor- tance of heat gains from lighting and solar radiation. Due to lower U-value materials in LEBs, it dampens the indoor temperature fluc- tuations throughout the day resulting in an equilibrium indoor climate. In addition to this, the lower U-value increases the ther- mal resistance of the building by resulting in a slower heat transfer between the walls and indoor, and introduces a large time constant. Because of a large time constant as well as large heat capacity in LEBs compared with conventional buildings, it retains thermal gain from past climatic changes. Therefore, an estimation and prediction of LEB’s thermal energy demand is quite challenging. There are various prediction models based on physical, semi- physical and data-driven methods available to estimate and predict the thermal energy demand for different energy standard build- ings. Physical methods estimate the energy demand of a building from known parameters, i.e., detailed geometrical information and thermal properties of the building. Several physical building sim- http://dx.doi.org/10.1016/j.enbuild.2016.11.009 0378-7788/© 2016 Elsevier B.V. All rights reserved.