Elsevier

Energy

Volume 97, 15 February 2016, Pages 506-516
Energy

Estimating building energy consumption using extreme learning machine method

https://doi.org/10.1016/j.energy.2015.11.037Get rights and content

Highlights

  • Buildings consume huge amounts of energy for operation.

  • Envelope materials and insulation influence building energy consumption.

  • Extreme learning machine is used to estimate energy usage of a sample building.

  • The key effective factors in this study are insulation thickness and K-value.

Abstract

The current energy requirements of buildings comprise a large percentage of the total energy consumed around the world. The demand of energy, as well as the construction materials used in buildings, are becoming increasingly problematic for the earth's sustainable future, and thus have led to alarming concern. The energy efficiency of buildings can be improved, and in order to do so, their operational energy usage should be estimated early in the design phase, so that buildings are as sustainable as possible. An early energy estimate can greatly help architects and engineers create sustainable structures. This study proposes a novel method to estimate building energy consumption based on the ELM (Extreme Learning Machine) method. This method is applied to building material thicknesses and their thermal insulation capability (K-value). For this purpose up to 180 simulations are carried out for different material thicknesses and insulation properties, using the EnergyPlus software application. The estimation and prediction obtained by the ELM model are compared with GP (genetic programming) and ANNs (artificial neural network) models for accuracy. The simulation results indicate that an improvement in predictive accuracy is achievable with the ELM approach in comparison with GP and ANN.

Introduction

Energy consumption has deep implications for the socio-economic and political spheres of countries. The uncontrolled use of natural resources and energy reserves, has had a tendency to lead to environmental damage that endangers all life, hence the importance of saving energy. Energy waste presents an environmental hazard that warrants global attention to sustainability. In fact, a good percentage of energy waste is connected with buildings [1]. In 2012 alone, roughly 40% of the total U.S. energy consumption was used in buildings, both commercial and residential, thus, energy conscious construction is important to secure a sustainable future for humanity and the earth. Both, the commercial and residential sectors need to focus greater attention to energy use in the near future [2]. The operational energy of a building's useful life constitutes the greater part of its total energy usage (assuming the lifespan of a building is 75 years) [3]. Energy needs arise in various building and environment constituents, where climatic conditions, choice of building envelope and building insulation usually head the list. The thermal homeostatic regulation of interior spaces is greatly influenced by walls, roofs and glazing. Thus, the type of building envelope and its properties determine heat loss, heat gain, and air entering from outdoors [3], [4]. Insulation is a major factor in helping steady interior temperature and reducing the energy needed for the acclimatization of heating and cooling. Insulating materials reduce thermal transmittance (U-value) in the building envelope components, which not only contributes in reducing the required HVAC system size but also reduces energy costs [5]. Hence, in both naturally ventilated and air-conditioned buildings, the heat storing capacity of wall and roof construction affect indoor thermal comfort and cooling loads.

It is essential today to understand thermal comfort under the light of energy efficiency [6], [7], [8], [9], [10]. The fulfillment of energy efficiency in buildings requires reduction of energy consumption during their operation years. This requirement highlights the importance of energy usage estimation prior to design and construction stages, which can lead to significant improvements in the new constructed buildings in terms of energy efficiency [11], [12], [13]. The reduction of energy usage by engineered materials and proper design of the physical structure is well documented in the literature. Other studies have centered on energy requirements during a building's operational life span [14], [15], [16], [17]. The energy requirement and environmental impact of construction, has also been investigated [14], [15], [16], [18], [19]. Some studies have focused on the particular effects of a building material on energy consumption and have forecasted energy consumption by other software and neural network applications [11], [12], [20], [21], [22]. Then, several commercial products have been introduced to improve energy efficiency in residential buildings [6], [7], [10], [58]. Some companies use lightweight frames with standardized sections and easily degradable and recyclable materials to reduce energy needs, construction waste and costs [23], [24], [25], [26]. One of these systems is the common wooden ‘platform framing’. This lightweight framing can also be built in light steel, and is used widely around the globe [27].

The main objective of this study is to estimate a building's energy consumption using the extreme learning machine method, which will be referred to as ELM in the remainder following pages. The input data in the ELM method represent wall material parameters, such as type of material, thickness and thermal properties. The output data come from the EnergyPlus simulation results for sample buildings of different input parameters.

The materials and systems focused in this study, are prevalent in Eastern Europe and Turkey [12]. Turkey is a country with rising significance in the world energy market, in which energy demands have increased sharply during past decades. It is expected that energy demands in this country will grow by 4.5% from 2015 to 2030 [28]. A significant percentage of total life expenses in Turkey are associated with energy costs since natural gas and crude oil are imported [29]. It is hence necessary to minimize the building energy consumption in this region.

So far there has been several approaches of computational intelligence methods for application in prediction of energy consumption in buildings and other systems. For example there has been an investigation of prediction power consumption of graphic processing units with fuzzy wavelet neural networks in which the average error of the model was around 6% [30]. In another investigation a comparison of integrated clustering methods for prediction of building energy consumption was performed [31]. It was found in this paper that there was an inherent tradeoff between prediction accuracy and cluster stability. Furthermore, a prediction of building energy consumption using real coded genetic algorithm has been carried out based on support vector machine approach [32]. It was confirmed that genetic algorithm approach is superior to conventional approaches. Therefore to improve the prediction of building energy consumption and improve accuracy there is a need to analyze more sophisticated methods.

The building energy-efficiency analysis needs accurate on-line identification to determine optimal energy consumption. In this study, we introduced an estimation model for energy consumption by using the soft computing approach of ELM. The application of this modern computational approach to determine optimal values and functions in real world problems, is receiving high attention from in various areas of science. Different engineering fields have been applied NN (neural network) as a chief computational platform. This method is able to solve complex nonlinear problems which are difficult to achieve using classic parametric methods. Various algorithms such as SVM (support vector machine), BP (back propagation), and HMM (hidden Markov model) can be used for training in neural network. The downside of NN is the time required to learn it. ELM was introduced by Huang et al. as an algorithm for single layer feed forward NN. This algorithm is capable of solving problems caused by gradient descent based algorithms like back propagation which is applicable in ANNs. Training time required for using NNs can be reduced by ELM and by utilizing this method the learning process is much faster. A number of investigations have been carried out related to the application of the ELM algorithm to successfully solve problems in various scientific fields [36].

In this study, an attempt was made to retrieve the correlation between the main building envelope's parameters and the district heating and cooling loads. To achieve this purpose, up to 180 simulations were used to generate an ELM Predictive model.

Section snippets

Sample building

A residential building plan is used in EnergyPlus simulations (Fig. 1). This plan is an example of a typical single family house located in Istanbul, Turkey, built in a lightweight wood frame structure. Different building envelope scenarios were applied to this home plan.

The architecture plans of ground floor and first floor are shown in Fig. 1. Also Table 1 provides some basic information about the same building.

Materials

Fig. 2 shows the detailed section of wall construction used in the sample

Energy consumption

Table 7 shows the results of EnergyPlus software for all 180 simulations. The results obtained from EnergyPlus software are annual energy use for total district heating and cooling for all five groups. Annual energy use for interior lighting and equipment are the same for all buildings as the default inputs for lighting and equipment are constant in all models. These values are eliminated in this evaluation. In Table 7 it can be seen that the total energy needed for heating and cooling of the

Conclusion

This study carried out a systematic methodology to create an ELM building energy consumption predictive model. This method is applicable in residential projects to save time in an energy simulation procedure. It seeks a solution to achieve a method that can forecast the energy consumption of buildings by entering input data as material properties and thicknesses. The outcome of this research indicate that the proposed model can efficiently predict the energy consumption of buildings regarding

Acknowledgment

This research work is funded by High Impact Research Grant (HIRG) no. UM.C/625/HIR/VC/206 (Synthesis of Energy Redeemable Material from Local Wastes for Building).

References (47)

  • K. Kaygusuz et al.

    Renewable energy and sustainable development in Turkey

    Renew Energy

    (2002)
  • Haifeng Wang et al.

    Predicting power consumption of GPUs with fuzzy wavelet neural networks

    Parallel Comput

    (May 2015)
  • David Hsu

    Comparison of integrated clustering methods for accurate and stable prediction of building energy consumption data

    Appl Energy

    (15 December 2015)
  • Hyun Chul Jung et al.

    Prediction of building energy consumption using an improved real coded genetic algorithm based least squares support vector machine approach

    Energy Build

    (1 March 2015)
  • X. Wang et al.

    Online sequential extreme learning machine with kernels for nonstationary time series prediction

    Neurocomputing

    (2014)
  • A.H. Neto et al.

    Comparison between detailed model simulation and artificial neural network for forecasting building energy consumption

    Energy Build

    (2008)
  • G.B. Huang et al.

    Extreme learning machine: theory and applications

    Neurocomputing

    (2006)
  • Energy Consumption by Sector, U.S. Energy Information Administration (EIA), Independent Statistics & Analysis,...
  • V. Bokalders et al.

    The whole building handbook-how to design healthy, efficient and sustainable buildings

    (2010)
  • M. Guertin

    Green applications for residential construction

    (2011)
  • Center for Sustainable Systems

    Residential buildings factsheet

    (2009)
  • W.P. Spence

    Residential framing: a homebuilder's construction guide

    (1993)
  • RSMEANS

    Residential & light commercial construction standards

    (2008)
  • Cited by (0)

    View full text