Paper Conference

Proceedings of Building Simulation 2021: 17th Conference of IBPSA


A simplified open-loop control strategy for integrated shading and lighting systems using machine learning

Jiarong Xie, Azadeh Omidfar Sawyer
Carnegie Mellon University, United States of America

Abstract: Conventional integrated shading and lighting systems are usually sensor-dependent, which could entail excessive cost and labor associated with sensor installation, calibration, and maintenance. Advanced systems use daylight modeling to eliminate the use of physical sensors. However, real-time daylight simulation can be computation-heavy, leading to a slow response of the system. This paper proposed a data-driven method for integrated shading and lighting control, employing machine learning models developed from pre-simulated data to predict real-time daylighting and control the blind and lighting accordingly. Verification using climatebased daylight simulation with a case study showed that the method prevented 94.7% of annual glare and reduced lighting use by 64%. The study will contribute to the development of effective daylight-linked control systems for industrial applications.
Keywords: Glare control, automated shading system, data-driven method, daylight simulation
Pages: 3172 - 3179