A Data Driven Approach for Smart Lighting

Abstract

Smart lighting for commercial buildings should consider both the overall energy usage and the occupants’ individual lighting preferences. This paper describes a study of using data mining techniques to attain this goal. The lighting application embraces the concept of Office Hotelling, where employees are not assigned permanent office spaces, but instead a temporary workplace is selected for each check-in staff. Specifically, taking check-in workers’ light requirements as inputs, a collective classification strategy was deployed, aiming at simultaneously predicting the dimming levels of the shared luminaries in an open office sharing light. This classification information, together with the energy usages for possible office plans, provides us with lighting scenarios that can both meet users’ lighting comfort and save energy consumption. We compare our approach with four other commonly used lighting control strategies. Our experimental study shows that the developed learning model can generate lighting policies that not only maximize the occupants’ lighting satisfaction, but also substantially improve energy savings. Importantly, our data driven method is able to create an optimal lighting scenario with execution time that is suitable for a real-time responding system.