Forecasting Energy Consumption in Residential Buildings using ARIMA Models

Forecasting energy consumption in residential buildings is an important information for energy providers to meet the demands of consumers and plan properly grid resources. Moreover, it can be useful to support residents to reduce their utility bills and save energy. In this paper, we presented preliminary short-term forecasting results by utilizing autoregressive integrated moving average (ARIMA) models. Our analysis also connects the forecasting to the number of residents, employment status, number of electrical appliances, and size of the house. We evaluated our model on a publicly available dataset and optimal parameters were obtained through grid search. The mean absolute percentage error (MAPE) is calculated to quantify the forecasting error (i.e., 17.25%). The obtained results confirmed the applicability of our model to real-life applications.