A spatio-temporal deep learning-based decision support system for energy awareness in IoT-based smart buildings

Abstract

The increasing demand for energy and rising expectations for user comfort necessitate the more accurate and efficient management of climate control systems in smart buildings. A crucial step in this process is reliably predicting indoor temperature. In this study, multivariate time series data, including environmental parameters such as temperature, Relative Humidity (RH), light, and Heating, Ventilating and Air Conditioning (HVAC) consumption, were used to evaluate the performance of various deep learning models. Hybrid approaches integrating Recurrent Neural Networks (RNN) architectures, including Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) models, with Graph Convolutional Networks (GCN)(GCN-RNN, GCN-LSTM, GCN-GRU) were systematically compared. Furthermore, the Transformer architecture and the Extreme Gradient Boosting (XGBoost) algorithm were included in the comparison as a baseline reference. The results show that the GCN-GRU model achieved superior accuracy compared to other models in the analyzed regions and throughout the test period, reaching an R2 score of 0.9976 with low error rates and providing consistent accuracy. Beyond model performance, a user-friendly interface has been developed that enables the selection of alternative models, interactive visualization of prediction results, examination of the impact of the current control strategy on energy efficiency, and dynamic integration of new algorithms, thanks to a modular software architecture. These findings emphasize the importance of jointly processing temporal and spatial patterns and provide a practical foundation for decision support systems aimed at enhancing energy awareness and operational sustainability in IoT-enabled smart buildings.