ScholarWorks@Y-Scholar Hub

초록

While artificial lighting significantly impacts occupants' health and productivity in commercial buildings, current fixed lighting conditions lack the adaptability to real-time occupant status, highlighting the need for a more responsive and data-driven approach. This study presents the Cognitive Performance (CP), Visual Fatigue (VF), and Energy consumption Optimization (CVEO) model for occupant-centric lighting control. The CVEO model aims to search for correlated color temperature and illuminance that maximize CP while minimizing VF and energy consumption using non-dominated sorting genetic algorithm-II. In this study, the objective functions involve stacking ensemble models that predict CP and VF based on occupants' bio-signals (i.e., heart rate variability, brain waves, and ocular response) and tree-based machine learning algorithms. For model evaluation, experiments were conducted based on the collected data from 15 subjects. As a result of evaluating stacking ensemble models, personal models outperformed general models, with root mean squared errors averaging 0.023 and 0.012 for CP and VF predictions, respectively. Additionally, in 2,000 scenarios with randomly determined presence of the 15 occupants, the CVEO model showed better performance in objectives than fixed lighting conditions from previous studies. The CVEO model can dynamically reflect real-time occupant status through bio-signals, thereby optimizing lighting to enhance work productivity, alleviate eye-related symptoms, and achieve energy savings. This approach offers significant potential for improving indoor environmental quality and occupants' well-being.