Predicting Adverse Thermal Events in Smart Buildings

This research focuses on the prediction of adverse thermal events in smart buildings by leveraging machine learning techniques and advanced statistical modeling. The study utilizes the Adverse Condition and Critical Event Prediction Toolbox (ACCEPT) framework developed by NASA, which is based on the Multivariate State Estimation Technique (MSET). The process includes training a model on normal operating data, monitoring new data to estimate expected values, and detecting faults using the Sequential Probability Ratio Test (SPRT). The implementation demonstrates the ability to accurately predict anomalies in thermal events, thus enhancing the operational efficiency and safety of smart building environments.