Occupant-Centric Shading and Lighting Control Using a Simulation-Assisted Data-Driven Framework

Automated shading systems play an important role in balancing the benefits and drawbacks of daylight ingress within indoor environments. When properly operated, these systems can significantly reduce building energy consumption and ensure adequate daylight ingress and view access without compromising occupants’ comfort. Maintaining occupants’ visual comfort is of utmost importance for the successful functioning of shading controls, as it prevents occupants from manually overriding the system. However, there is currently a lack of simplified and practical methods to integrate the concept of visual comfort and occupants’ preferences into the shading control process. Consequently, current automated shading control is often overridden by occupants and fails to balance the advantages and disadvantages of daylight ingress. To address these challenges, this dissertation aims to propose simplified and practical occupant-centric methods for integrated shading and lighting control, to promote the application of automated shading in real-world settings. 

The primary objective of this dissertation is to introduce an occupant-centric shading and lighting control framework designed to effectively address the issue of daylight glare. This objective was achieved with machine learning models developed using pre-simulated data. These models require a minimal set of input variables and deliver rapid, accurate glare predictions. The implementation and effectiveness of the framework were validated through two case studies using climate-based daylight simulations. The results suggest that the proposed control framework could eliminate between 86.5% and 96.9% of annual glare and reduce lighting energy use by 80.8% when preventing glare for an individual occupant. When controlling glare for multiple occupants, the control framework could prevent glare up to 97% of the time while decreasing lighting energy use by 19.4%. The second case study further emphasized the significance of algorithm selection, strongly favoring Random Forest models and discouraging K-Nearest Neighbors and Support Vector Machine models.

Furthermore, this dissertation proposed an “open-only” approach to include occupants in the control loop by leaving the closing actions to users, aiming to improve their satisfaction with automated shading systems. Human subject experiments were conducted to compare this new approach against the conventional “open + close” control. The participants’ responses indicated a clear preference for, and greater satisfaction with, the “open-only” control method. Additionally, an analysis of measured work plane illuminance suggested that the proposed control approach, when integrated with lighting control, could potentially decrease lighting energy consumption by 33.0%. 

To conclude, the simulation-assisted data-driven control framework eliminates the use of indoor camera sensors or real-time computationally intensive daylight modeling. This approach provides a simplified and practical method to incorporate glare considerations into the control of shading systems, which could significantly improve occupants’ visual comfort and satisfaction with the shading system. The presented “open-only” control approach could further enhance user satisfaction by including occupants in the control loop in a practical manner. Both methods contribute to a practical user-centric shading control solution for reallife applications.