Campus Energy usage modelling for a South African Academic Institution

When considering buildings of academic institutions, various drivers often impact the building's energy consumption. This paper evaluates the influence of average temperature, work days, non-work days, and monthly active hours towards the energy usage of selected campus buildings. Various existing models are utilized to evaluate the sensitivity and influence of each of these drivers on the buildings energy usage. These models include: linear two variable and multivariable regression, exponential regression and polynomial regression. The dataset selected consists of energy data from selected campus buildings of a South African University, during the 2017 academic year. The buildings were clustered into three categories, namely libraries, student centers and laboratories. The results of this study revealed that the models developed using polynomial regression produced coefficient of determination (R²) values ranging from 0.48 to 0.9 in the case of active monthly hours, 0.32-0.70 in the case of average temperature and 0.21-0.8 in the case of workdays and non-workdays, which were the highest model accuracies when compared to those of other models. Analysis of the results also revealed that certain drivers such as active monthly hours had a greater correlation to the building energy usage over other drivers. The final model developed (A linear multivariable regression model) achieved an R2 value of: 0.9 (student centers), 0.85 (libraries) and 0.89 (laboratories), indicating the models high accuracy in predicting the dependent variable (energy consumption) when using all the drivers indicated as independent variables to the model.