Performance analyses of statistical approaches for modeling electricity consumption of a commercial building in France

Abstract

Energy consumption of buildings is an important component of national energy consumption, and, thus, developing accurate models for predicting future demand in buildings is essential for facility managers, electricity providers, and policy makers in tackling the energy scarcity. In this study, multiple linear regression (MLR), time series and Grey models were developed for estimating the HVAC electricity consumption of a commercial building located in Paris, France. The data was collected between June 2015 and April 2016. Weather variables (outdoor temperature, relative humidity, global radiation) and four dummy variables, which represent the working days, were taken into account in the regression analysis. Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Mean Squared Error (MSE) were selected for comparing the performance of the recommended models. The results show that MLR performs better with a RMSE of 18.3806 compared to time series and Grey model with RMSE of 20.5114 and 21.8478, respectively.