Average modelling of medium frequency DC/DC converters in dynamic studies

This paper studies the average value modelling (AVM) of medium frequency (200 Hz-2000 Hz) dc-dc converters, which are becoming increasingly important in high-power dc grids and distributing microgrids. The most convenient modelling is based on transfer function and power components which are available in common simulation platforms, such as PSCAD/EMTDC. However, these models are found to require very small simulation steps and the achieved time-domain accuracy is not satisfactory for medium frequency converters. The core issue of accuracy loss is identified and corresponding improvement is proposed. User written state-space models in the abc frame and the dq frame are analyzed in depth. Different solvers, including the Dommel's method, the Heun's method, and the Runge-Kutta method were adopted and compared in solving the state-space models. A high-power 2-kHz LCL dc-dc converter is taken as the test system. It is found that the AVM in the dq rotating frame based on the Runge-Kutta solver is able to generate very accurate time-domain responses compared to the detailed switching model with almost 100 times improvement in simulation speed. It is demonstrated that a dc-dc converter with 2-kHz inner ac circuit frequency can be accurately simulated at a 50- μs time step.