This paper introduces an innovative methodology for thermal comfort measurement based on a literature validated comfort model, namely the simplified Predicted Mean Vote (sPMV). The methodology is applied in a real-life case study, utilizing collected environmental data (e.g., indoor and outdoor temperature, and relative humidity). Aligned with recent advancements in the Built Environment (BE) digitalization, the study responds to the increasing demand of measuring comfort for improving the energy performance and the occupants’ well-being in the BE. The analyzed case study focuses on an office building equipped with a sensor network with a 1-year monitoring period. This research develops a LongShort Term Memory (LSTM) Network employing the measured environmental parameters. The network predicts comfort in terms of sPMV for two distinct forecasting horizons and its performance is evaluated by the Mean Square Error (MSE) and the Mean Absolute Error (MAE). The obtained validation metrics (MSE= 0.0095 , MAE= 0.026) outline the effectiveness of the proposed methodology, providing a reliable foundation for future developments in the domain of data-driven services in the BE.
AI-Based Methodology for Thermal Comfort Measurement: Application of a Simplified Comfort Model on a Real-Life Case Study
Pracucci, Alessandro;
2024
Abstract
This paper introduces an innovative methodology for thermal comfort measurement based on a literature validated comfort model, namely the simplified Predicted Mean Vote (sPMV). The methodology is applied in a real-life case study, utilizing collected environmental data (e.g., indoor and outdoor temperature, and relative humidity). Aligned with recent advancements in the Built Environment (BE) digitalization, the study responds to the increasing demand of measuring comfort for improving the energy performance and the occupants’ well-being in the BE. The analyzed case study focuses on an office building equipped with a sensor network with a 1-year monitoring period. This research develops a LongShort Term Memory (LSTM) Network employing the measured environmental parameters. The network predicts comfort in terms of sPMV for two distinct forecasting horizons and its performance is evaluated by the Mean Square Error (MSE) and the Mean Absolute Error (MAE). The obtained validation metrics (MSE= 0.0095 , MAE= 0.026) outline the effectiveness of the proposed methodology, providing a reliable foundation for future developments in the domain of data-driven services in the BE.I documenti in SFERA sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
