Sensorless Illumination Control of a Networked LED-Lighting System Using Feedforward Neural Network

In order to resolve the problem of energy hunger nowadays, saving lighting energy in buildings contributes an important part. In this paper, a sensorless illumination control scheme for smart networked LED lighting has been investigated. The scheme is based on a feedforward neural network to model all the nonlinear and linear relationships inside the lighting system as the controlled plant. Because the scheme does not rely on lighting simulation software, it is flexible to be implemented on microcontrollers. The scheme, moreover, can provide not only high accuracy in modeling but also global optimum in energy saving. Without using light sensors in its control loop, the approach can save significant cost and provide ease of installation as well. In addition, it also has the strength of fast response owing to feedforward control based on neural networks. The experimental results show that the approach can easily attain more than 95% modeling accuracy and also improve more than 28% energy saving with its optimal nonlinear multiple-input multiple-output control.