Prediction of building energy consumption by using artificial neural networks

Abstract

In this study, the main objective is to predict buildings energy needs benefitting from orientation, insulation thickness and transparency ratio by using artificial neural networks. A backpropagation neural network has been preferred and the data have been presented to network by being normalized. The numerical applications were carried out with finite difference approach for brick walls with and without insulation of transient state one-dimensional heat conduction. Three different building samples with different form factors (FF) were selected. For each building samples 0–2.5–5–10–15 cm insulations are assumed to be applied. Orientation angles of the samples varied from 0° to 80° and the transparency ratios were chosen as 15–20–25%. A computer program written in FORTRAN was used for the calculations of energy demand and ANN toolbox of MATLAB is used for predictions. As a conclusion; when the calculated values compared with the outputs of the network, it is proven that ANN gives satisfactory results with deviation of 3.43% and successful prediction rate of 94.8–98.5%.

Introduction

After the crisis in 1970s, energy becomes one of the major topics in agendas of developed and developing countries which provide their energy needs from others [1]. It is clear that in view of the technological developments, it will continue to keep its place there. In Turkey, 80% of the consumed energy in houses is used up for thermal comfort, especially heating [2]. This amount of energy is luxury for countries like Turkey that import most of the energy they consume. Studies about minimizing energy consumption and using renewable energy sources speeded up with the reduction of fossil fuels and increasing of various environmental problems. Turkey has a very big potential to benefit from renewable energy sources. In spite of the absence of the needed substructure benefitting from these energy sources like sun, wind, water, etc. and the expense of the investment costs, there are only a few applied projects in Turkey. If the lack of technology and materiality are taken into consideration, the wisest solution is to minimize the extravagant consumption with energy efficient buildings.

The term energy efficiency in buildings can be defined as, providing comfort conditions by not sacrificing the indoor’s quality with minimum energy consumption. The necessity of replacing the lost or excessive gains in order to protect the indoor comfort conditions makes it necessary to use energy in buildings. The energy load of a building is the amounts of heating or cooling energy that must be taken on by heating and cooling systems. The energy efficiency in buildings will be achieved by ruling the heat transitions through the envelope components like walls, windows, roof, etc. [3]. Designing of the envelope as a layer which provides balance between indoor and outdoor will increase the efficiency. So, the temperature difference between inner and outer surfaces of the elements and heat transitions to outdoor will decrease and efficiency of the envelope will be accomplished [3]. Heat losses through the envelope can be controlled by facade system, materials, orientation and form of the building. Extra loads on buildings mechanical installation due to the usage of the inappropriate materials and orientation cause increasing of the life-cycle costs and appearance of the disturbances named sick building syndrome for users. These situations make it obligatory to use the present sources efficiently, reducing the environmental pollution caused by the excessive consumption of fuels and increasing the productivity of the users and decreasing the life-cycle costs.

There are many factors that affect the energy needs of buildings. Basically these factors can be divided into two groups as physical environmental and artificial designing parameters. The physical environmental parameters are; outdoor temperature (°C), the amount of the solar radiation (W), wind speed (m/sn), etc. and the artificial designing parameters are; transparency ratio (%), building form factor, orientation, optical and thermo-physical properties of the materials used in building envelope and the distance between buildings [4]. The transparency ratio is the percent area of a wall covered by windows while the form factor of a rectangular shaped building can be explained as the ratio of the areas of south and north facades to east and west facade areas. The orientation of a building is evaluated with surface azimuth angle (γ) in this study.

Artificial intelligence can be defined as the learning, understanding and thinking ability of computers. But the only thing that computers can do with today’s technology is to perform the algorithms which are given to them [5]. Artificial neural network (ANN) is an important branch of artificial intelligence and widely being used in many engineering problems.

In literature about ANN, Keleşoğlu and Fırat [6] have calculated the heat losses in brick wall and installation pipe and obtained satisfactory results. Yang et al. [7] have investigated the adaptation of the network to unexpected changes of the input variables and the ability to predict the on-line building energy needs. Kalogirou and Bojic [8] used ANN for the prediction of energy consumption of a passive solar building. As a result they proved that ANN gives quick results than dynamic simulation tools. Gonzalez and Zammareno [9] brought up the results of an ANN based method that predicts the electric loads of buildings. Olofson and Anderson [10] have developed a neural network which makes long-term energy demand predictions based on short-term measured data with high prediction rate of 90–95%. Ben-Nakhi and Mahmoud [11] have investigated the feasibility of the usage of neural network in optimizing the thermal energy storage of HVAC systems in public especially office buildings. And they found out that ANN is a strong tool in the optimization of thermal energy storage of buildings. Ayata et al. [12] have used ANN in prediction of the temperature distribution of the multi-layered metal plates by training with the numerical results of the problem obtained with finite elements methods and have achieved strong linear association between the predicted and calculated results with the ratio of 99%. Qi et al. [13] combine the ANN method with the conventional method and present the model of shower cooling tower with accuracy and adaptively. Chau [14], [15] proven that ANN can be successfully used in whole area predictions after a learning process and in his another study, he proven the usability of particle swarm optimization training algorithm as an alternative training algorithm for ANN.

This study is significant for, it gives way to energy efficient buildings in the designing stage and will be helpful for architects and design engineers in decreasing the energy consumption of new constructed buildings. Because the first precautionary measures, must be taken in designing period of buildings, against high energy costs and environmental problems caused by hazardous wastes. The main objective is to present the results of a method for prediction of heating loads of buildings with high accuracy by using backpropagation neural network and to compare the numerical results of conventional calculation method with ANN. In order to confirm the accuracy, three different networks were prepared for three different building samples. Each networks are employed to predict the heating energy needs of building samples having different transparency ratios, insulation thicknesses and orientations. The training and testing datasets were prepared with calculations of heating energy demands of buildings by using finite difference approach of transient state one-dimensional heat conduction problem [16]. The calculations of energy demands of the building samples were carried out with a computer program written in FORTRAN and ANN toolbox of MATLAB is used for the predictions.