Design and Implementation of Efficient Smart Lighting Control System with Learning Capability for Dynamic Indoor Applications

Abstract

Accurate and efficient adjustment of luminaire's dimming level in a smart environment can be a huge challenge. Indoor lighting system as a nonlinear and time variant block, which consumes significant amount of electrical power is evaluated in this paper. In doing so, a control method is proposed to efficiently adjust luminaire's dimming level in a smart environment and to optimize energy and user's comfort level. The proposed control method takes advantages from neural network and its learning capabilities. In this research, photodetectors are placed at the work zones, where work zones can have different number of photodetectors without any increase in complexity and any adverse effect on the control system. The method is capable of adopting itself to daylight variations with high accuracy. A state machine is developed to implement the method. The method is implemented in MATLAB and lighting conditions are extracted in DIALux. Luminaire's dimming levels are determined with accuracy higher than 99%. Daylight is considered as a bias to the system and thus the network does not need to be trained by any variations. In a dynamic condition, when taking into account the variation in daylight, the system mean error does not exceed 3%.

Machine Learning Methods in Smart Lighting Toward Achieving User Comfort: A Survey

“…Meanwhile, complex controls such as multi-color lighting control need intelligent methods such as artificial neural networks (ANN). The ANN method was also in use in a complex problem of the research by Seyedolhosseini et al [35]. The ANN method controls luminaire dimming with system faults and daylight variation detection without increasing complexity.…”