Analysis and partitioning of heterogeneous models of hybrid power trains with regard to real-time simulation

In this paper, an approach for the efficient handling of mathematically stiff models in the case of a real-time simulation is shown. Here, the dynamic characteristics of models created in the simulation environment "Dymola" are analysed and the sources of the model stiffness are identified. Based on the gained information and according to the constraints of the real-time simulation, the model is adjusted automatically. In a first step, based on the eigenvalue distribution in the region of stability the model is partitioned into several sub-models. Each of these submodels is homogeneous and not stiff. If one or more models violate the real-time constraints and still threaten the stability of the simulation, an automatic optimisation of the model parameters will be performed in a second step. This affects the model's time constants and therefore the dynamic behaviour of the system. Based on this methodology stiff models are automatically adjusted and optimised in terms of real-time simulation. Thus, long analysis phases and the high number of iterations to identify and to eliminate the model stiffness can be reduced. Therefore, the transition from modelling (MIL) to real-time simulation (HIL, SIL) becomes more efficient.