Daylighting and Glare Optimization in University Classrooms Based on Parametric Simulation and Machine Learning: A Case Study of Yunnan University

Abstract

Abstract

Low-latitude plateau classrooms, such as those in Kunming, experience intense solar radiation that often causes insufficient far-window illumination, excessive near-window brightness, and viewing-direction glare under side-lighting conditions. To investigate this spatial imbalance, field surveys of nine classrooms were used to define realistic parameter ranges, while 100 parametric design cases were evaluated through annual simulation, machine learning, and SHAP analysis. The viewing-direction glare model achieved a test-set R2 of 0.819, indicating adequate predictive performance for factor interpretation. Compared with simply increasing the window-to-wall ratio (WWR), coordinated control of classroom geometry, window configuration, and surface reflectance produced a more balanced luminous environment. Daylight availability and excessive illuminance were primarily governed by WWR and window reveal depth, whereas glare was more strongly influenced by seating position, viewing direction, window width, and orientation. Classroom-wide averages may therefore conceal localized glare experienced by students. A moderate WWR of 0.26–0.40 combined with a window reveal depth of 0.75–1.17 m emerged as a preferable strategy within the investigated design space. These findings support desktop-level daylight assessment and student-perspective glare evaluation in the design and renewal of ordinary side-lit classrooms in Kunming and comparable low-latitude plateau regions.