Abstract Large buildings are currently one of the top electrical energy consumers, accounting for 30% of global energy consumption and contributing to 26% of energy-related emissions. Accurate electricity demand forecasting supports energy management processes, enabling efficient energy consumption in large buildings, which are essential against climate change. Different tools have been used to forecast electricity demand over different building types, including statistical methods and neural network architectures. This study proposes a deep neural network model to forecast electricity demand in a university campus in Ecuador using multivariate time series data, including electricity demand, meteorological, and time-related variables. We evaluated four models: a statistical model (SARIMAX), along with three deep neural network models (Recurrent Neural Network (RNN), Gated Recurrent Unit (GRU), and Long Short-Term Memory (LSTM)) through four forecasting horizons (24-hour, 48-hour, 7-day, and 30-day). Findings show the efficacy of deep neural networks in forecasting the campus’s electricity demand, with high performance over all evaluated time horizons. Among these networks, the LSTM demonstrates superior performance, particularly over the 7-day and 30-day horizons. This study also seeks to explore how much historical data is necessary to train a well-performing model across these forecasting horizons.
Lugmania et al. (Tue,) studied this question.