Manchester Metropolitan University's Research Repository

Daylight adaptive smart indoor lighting control method using artificial neural networks

Seyedolhosseini, Atefesadat, Masoumi, Nasser, Modarressi, Mehdi and Karimian, Noushin (2019) Daylight adaptive smart indoor lighting control method using artificial neural networks. Journal of Building Engineering, 29. p. 101141. ISSN 2352-7102

Accepted Version
Available under License Creative Commons Attribution Non-commercial No Derivatives.

Download (1MB) | Preview


Accurate and efficient adjustment of maintained illuminance and illuminance uniformity in indoor environments with daylight variations is a tremendous challenge, mainly due to the nonlinear and time-variant nature of lighting control systems. In this paper, we propose a smart lighting control method for indoor environments with both dimmable (controllable) and uncontrollable external light sources. Targeting an indoor environment with multiple zones, each requiring a different lighting condition and equipped with an unequal number of photodetectors and dimmable light sources, this paper presents a novel control mechanism that determines the output flux of each luminary in such a way that each zone (1) receives the required maintained illuminance, (2) illuminance uniformity conditions are met inside each zone, and (3) the power consumption is optimized. This method uses a neural network to learn the impact of each luminary on the maintained illuminance of each zone and adjust the dimming level of the luminaries to establish the required illuminance in the zones. We also rely on photodetectors to measure the daylight illuminance continuously and use it as the bias value for the neural network. The new priority value allows losing some illuminance accuracy (by allowing lager difference between the actual and required maintained illuminance values) for low-priority zones to reduce power consumption. The method has been evaluated in different test cases by chaining the widely-used DIALux tool and some MATLAB toolboxes. The evaluation results show that the method can achieve considerable accuracy by yielding an average Mean Square Error of 1.2 between the demanded and sensed illuminance values. Furthermore, when all sensors except one reference sensor are removed from each zone (to increase user comfort or reduce cost), the mean square error is less than 25.4 across all considered test cases.

Impact and Reach


Activity Overview
6 month trend
6 month trend

Additional statistics for this dataset are available via IRStats2.


Actions (login required)

View Item View Item