Building dynamic models of LED lamps by system identification methods

This paper presents a system identification based modeling method for Lighting-emitting diode (LED) luminaires. The dynamics of luminaires in response to control inputs are considered in the identified models. For the system identification, a set of pseudorandom binary control signals are used as dimming inputs to excite the luminaire. Using the sampled light output from the luminaire, and by subspace identification, its dynamic model in state space form can be identified. Then the estimated lumen output can be calculated with Kalman filtering. Experiments on a real LED system have been conducted. The measurements and results confirm the validity of the dynamic modeling, and comparison with static linear modeling shows its advantage on prediction precision. The potential applications of the identified models are to design controllers and diagnostic observers for LED luminaires.