Neural network-based LED lighting control with modeling uncertainty and daylight disturbance

Abstract

This paper presents a neural network-based control method for achieving desired lighting levels in an LED-based lighting system with unknown or uncertain system model parameters in the presence of daylight disturbances. Assuming an unknown system model matrix, the control strategy utilizes an online neural network method to synthesize a learning controller. The control commands are dimming levels, which represent the percentage of LED's full power level and the outputs are illuminance levels at target points. The neural controller is designed using a Lyapunov-based analysis to achieve boundedness of the output error to an arbitrarily small ultimate bound. By considering the daylight as a disturbance, it serves as a bias for the desired control system set-point resulting in lower dimming command inputs and energy savings. The controller design only requires the tracking of error signal as input which eliminates the need for any prior knowledge of the daylight disturbance and room model.