Influence of building parameters on energy efficiency levels: a Bayesian network study

Design parameters of a building play a major role in its energy consumption. Towards this, we studied the energy efficiency of buildings using the association and dependence of input variables, such as relative compactness, surface area, wall area, roof area, overall height, orientation, glazing area and glazing area distribution, to the output variables-heating load (HL) and cooling load (CL). Bayesian network, a supervised machine learning model, was used to identify dependencies between variables. UCI energy efficiency dataset (768) with eight-labelled inputs was used to make predictions with 10-fold cross validation. The Bayesian network was chosen to identify the most impactful input parameters. Seven search algorithms to determine the Bayesian network structure based on training data were considered to analyze the best-performing algorithm for predicting the relationship between nodes. Among those, Tabu search (82.81% and 81.77%) and Simulated annealing (82.68% and 81.38%) performed best with highest accuracies for both HL and CL. In addition, it is found that reduced heights of buildings will have a very high-energy efficiency level for both HL and CL. Reduced glazing areas will have a high-energy efficiency level for HL. These findings could be used to build real-world higher energy efficient structures.