Integrating Experimental Analysis and Machine Learning for Enhancing Energy Efficiency and Indoor Air Quality in Educational Buildings

Ensuring energy efficiency and maintaining optimal indoor air quality (IAQ) in educational environments is vital for occupant health and sustainability. This study addresses the challenge of balancing energy consumption with IAQ through experimental analysis integrated with advanced machine learning (ML) techniques. Traditional methods often fail to optimise both simultaneously, necessitating innovative solutions leveraging real-time data and predictive models. The research employs ML models, including Recurrent Neural Networks (RNN), Long Short-Term Memory Networks (LSTM), Gated Recurrent Units (GRU), and Convolutional Neural Networks (CNN), using a dataset of over 35,000 records. Parameters such as CO2 levels, particulate matter (PM), temperature, humidity, and exogenous variables (e.g., time, date, and rain sensor) were analysed to identify environmental factors influencing HVAC system efficiency. Predictive models achieved over 92% accuracy, enabling precise real-time HVAC control to balance energy use and IAQ. Key findings highlight GRU and LSTM models' effectiveness, with scalability across educational institutions showing potential for reducing energy costs and improving indoor environments. Validation with diverse datasets demonstrated robustness, while SHAP (Shapley Additive exPlanations) values provided enhanced interpretability, helping policymakers and managers implement effective strategies. This research underscores the transformative role of ML in optimising HVAC efficiency and IAQ management, offering scalable, data-driven strategies to reduce carbon footprints, improve occupant well-being, and align with global sustainability goals. By overcoming traditional limitations, the study presents a systematic approach for integrating empirical data with AI, advancing smarter, healthier, and more sustainable learning environments.