How to reduce the simulation duration driving cycle simulation with high time span?

Question

Hello everyone,

I´m actually working on a simulation model of an electric power train of a battery electric bus.
The main components are a HV-Battery, a DCDC Converter and a induction machine with the inverter for the control of speed and torque.
With the simulation model the behaviour of the components of the power train should be evaluated by NEFZ and WLTP driving cycle. So the time span of the simulation between 1200s and 1800s.

I tested my simulation model with a shorter time span of 150s and it already took over 8h to finish the simulation.
For speed up the simulation i reduced the scopes to a minimum of 7 and raised the maximum step size.
The sample time of the scopes i also raised, but it takes still very long time for the simulation even with a powerfull i7 processor in the pc where Plecs V4.5.8 is running.
Which options do I have to increase the simulation speed significantly?
For a better support I attached my simulation model.

Answer 1

Running long duration simulations with relatively complicated models will be time consuming.  There often isn't one simple fix to make drastic improvements, but a a few suggestions are listed below:

• In your simulation settings I noted that the maximum number of consecutive zero crossings is set to a very large value.  Did you encounter an error previously? If a scenario exists where there are many consecutive zero crossings, your simulation will proceed extremely slowly.  You should resolve this first by understanding the source of the error.
• As you pointed out, for longer simulation runs storing and displaying scope data can be very memory intensive.  You should look into reducing the total memory required by reducing the number of scopes and traces per scope.  One can also set a sample limit and set a sample time in the "Scope parameters" settings associated with each scope.  The To-File block may also be helpful.
• For longer simulations, consider if an average model would be suitable. This would enable the simulation to run faster, but may not be useful in all scenarios.  For example, you cannot use thermal models with average switch models. 
• In general, consider the model complexity - does one need such a detailed battery model? Are there any small time constants in the model that would drive a very small time-step? For any hysteretic controls, one should understand if the hysteresis bands are appropriately set.