ITempDT: AI-Powered Digital Twin for Forecasting Indoor Temperature in Smart Buildings

This work presents iTempDT, an AI-driven Digital Twin (DT) approach that uses deep learning to forecast indoor temperature based on Heating, Ventilation, and Air Conditioning (HVAC) system operations. Accurate temperature forecasting is essential for smart buildings (SBs), which aim to optimize energy efficiency without compromising occupant comfort. iTempDT addresses this need by providing accurate and timely forecasts that allow intelligent HVAC control and energy-efficient decision-making. At the core of iTempDT is a Temporal Convolutional Network (TCN) integrated with an attention mechanism to capture temporal dynamics and feature dependencies from multivariate time series data collected from both indoor and outdoor sensors. The model is trained for multi-step indoor temperature forecasting, explicitly conditioned on the recent HVAC status (ON/OFF), allowing for real-time predictions as well as scenario-based simulations. Also, an interactive dashboard enables real-time visualization and virtual control of HVAC settings. To promote sustainable and energy-conscious AI practices, iTempDT integrates a Green AI approach by applying post-training quantization. This reduces the model size and speeds up inference, making it suitable for deployment on low-power edge devices. The DT prototype was developed at the ICAR-CNR IoT Lab in Rende, Italy. iTempDT shows strong performance across key metrics (MAE, R2, and RPD). Finally, the proposed approach is compared with other state-of-the-art models.