Dynamic compact thermal model extraction for LED packages using model order reduction techniques

Thermal management is one of the key issues arising in designing light-emitting diode (LED) based luminaire products. Dynamic compact thermal models (DCTMs) are required to predict the thermal behaviour of LED packages fast and accurately in system level simulations. The European Delphi4LED consortium aims to develop multi-domain (electrical-thermal-optical) compact models. One of its targets is to develop a methodology to extract DCTMs that can handle multiple heat sources of LEDs. We present the implementation of Krylov subspace based model order reduction techniques to extract the DCTMs of LED packages selected in the Delphi4LED simulation benchmarks. The results presented in this paper show that the extraction of DCTMs using iterative rational Krylov is highly accurate for these benchmark problems.