The startup and shutdown simulation of a micro gas turbine (MGT) is concerned. In order to build an accurate dynamic model, the general models of core components are presented first. Then, two dynamic parameters that influence rotating speed and temperature dynamics are put forward for further identification. The identification is accomplished by a reinforcement learning algorithm. The comparison between the simulation and experimental data show that this learning method leads to high-accuracy MGT startup and shutdown simulation.
Keywords micro gas turbine, reinforcement learning, dynamic modeling, startup and shutdown simulation, distributed energy generation