| Title |
Simulation-assisted Optimal Lighting Control for a Factory Building |
| Authors |
김영섭(Kim, Young-Sub) ; 김재민(Kim, Jae Min) ; 신한솔(Shin, Han Sol) ; 박철수(Park, Cheol-Soo) |
| DOI |
https://doi.org/10.5659/JAIK.2020.36.8.127 |
| Keywords |
daylighting; electric lighting; lighting control; lighting simulation; DIVA; machine learning; surrogate model |
| Abstract |
Lighting control can be categorized into open-loop and closed-loop. For the closed-loop control, illuminance sensors are generally mounted on
a horizontal workplane, or the nearest wall/ceiling. As the size and complexity of an indoor space increases, the number of sensors and its
corresponding control become complex because illuminance at a certain point is influenced by multiple neighboring lighting fixtures. The
open-loop control is disadvantageous because it can’t reflect the illuminance level of a workplane. With this in mind, the authors aim to
develop an approach where lighting simulation model could predict the illuminance level at any points of interest, hereby replacing
illuminance sensors, and lead to electric lighting energy savings. For this purpose, Radiance, one of the most sophisticated lighting simulation
tools, was first employed for daylit and electric lighting prediction of a target building. Then, a surrogate model, ANN (Artificial Neural
Network) model, was developed for fast computation and optimal control. Unknown parameters, e.g. reflectances of ceiling, floor, walls,
transmittance of glass and light loss factor, were estimated. It was found that the calibrated model’s prediction is accurate and the proposed
approach can save lighting energy by 18.6% for three days’ validation period (Mar 9-12, 2020) conducted at the target building. |