Prediction and Analysis of Building Energy Efficiency Using Artificial Neural Network and Design of Experiments

Abstract

Energy consumption of buildings is increasing steadily and occupying approximately 30-40% of total energy use. It is important to predict heating and cooling loads of a building in the initial stage of design to find out optimal solutions among various design options, as well as in the operating stage after the building has been completed for energy efficient operation. In this paper, an artificial neural network model has been developed to predict heating and cooling loads of a building based on simulation data for building energy performance. The input variables include relative compactness, surface area, wall area, roof area, overall height, orientation, glazing area, and glazing area distribution of a building, and the output variables include heating load (HL) and cooling load (CL) of the building. The simulation data used for training are the data published in the literature for various 768 residential buildings. ANNs have a merit in estimating output values for given input values satisfactorily, but it has a limitation in acquiring the effects of input variables individually. In order to analyze the effects of the variables, we used a method for design of experiment and conducted ANOVA analysis. The sensitivities of individual variables have been investigated and the most energy efficient solution has been estimated under given conditions. Discussions are included in the paper regarding the variables affecting heating load and cooling load significantly and the effects on heating and cooling loads of residential buildings.

An Octahedric Regression Model of Energy Efficiency on Residential Buildings

“…Artificial Neural Networks (ANN) [2,[9][10][11][12][13][14][15][16][17][18][19]] Decision Trees (DT) [20] Support Vector Regression (SVR) [11,13,14,19,[21][22][23]] Random Forest (RF) and Trees Ensemble [8,13,14,[20][21][22][23][24][25][26][27][28][29] Multi-Layer Perceptron (MLP) [14,20,23,27] Gaussian Mixture Model (GMM) [30] Gradient Boosted Regression Trees (GBRT) [24,31] Extreme Learning Machine (ELM) [10,32,33] Linear Regression (LR) [10,13,21,23,27,34,35] Radial Basis ...…”

“…In reference [18], the considered variables are (surface area, wall area, roof area, overall height, glazing area), while reference [35] uses (relative compactness, surface area, wall area, overall height, glazing area) to model the cooling problem. The importance of each variable is studied in [9] using ANOVA with the result of (relative compactness, surface area, wall area, overall height, glazing area) and (relative compactness, wall area, roof area, overall height, glazing area) as the most important variables for the heating and cooling problems, respectively.…”

“…Sci. 2019,9, x FOR PEER REVIEW 16 Relationship between the independent variables for the cooling load problem. (a) Surface vs. Compactness; (b) Wall vs. Compactness; (c) Roof vs. Compactness; (d) Height vs. Compactness; (e) Orientation vs. Compactness; (f) Glazing area vs. Compactness; (g) Glazing distribution vs. Compactness; (h) Wall vs.…”

Estimation of Building Energy Efficiency Performance Using Radial Basis Function Neural Network

“…Heating load (HL) and cooling load (CL) contribute around 30-40% of building's energy usage. Minimizing the HL and CL plays a major role in ensuring total energy consumption reduction in buildings [2]. Previous research in building energy based on computer experiment is done to optimize parameters for energy efficiency improvement in the residential or commercial buildings.…”

“…Recent studies in computer simulation of energy consumption have been conducted using random forest algorithm (RF) [3], support vector machine (SVM), polynomial regression, decision trees, artificial neural network (ANN) [2], genetic programming (GP) [4], support vector regression (SVR) with ANN [5] and multilayer perceptron (MLP) with grid search strategy algorithm [6]. Major study in energy building efficiency by T.Sanas and Xifara (2012) identified that RF hugely outperformed IRLS in finding an accurate functional relationship between the input and output variable.…”