Data-Driven Energy-Efficiency and Comfort Optimisation in Indoor Environments

Climate change, uncertainties in energy prices, and the Covid-19 pandemic have significantly reshaped building management, highlighting the need for energy-efficient, safe, and comfortable indoor environments. With advancements in Internet of Things (IoT) sensors and Artificial Intelligence (AI) techniques, optimising building performance now includes forecasting key parameters and intelligently controlling Heating, Ventilation and Air Conditioning (HVAC) systems. However, existing studies often lack practical applicability in real-world scenarios, typically relying on extensive data collection or tailored physical/mathematical models, with limited focus on deployment, scalability, and long-term performance. This thesis addresses the problem from a different angle, proposing an adaptive and practical AI-based solution for energy-efficient comfort optimisation in indoor environments. The designed approach continuously learns from the monitored environment through collected data and requires minimal human effort for configuration and maintenance. The contributions are as follows: i) a method for accurately predicting key parameters using a limited window of data, with a dynamic mechanism to keep the AI model current with environmental changes and operational in a short time frame, and ii) a novel algorithm called EECO for automated and intelligent HVAC control, driven by continuous short-term decisions based on long-term predictions to balance thermal comfort and energy consumption, with no need for preliminary knowledge of the local environment. Evaluation results demonstrate that the proposed approach achieves high prediction accuracy, ensures desired thermal comfort, and reduces the energy footprint by up to approximately 16% in a real-world environment, in addition to potentially saving on operating costs.