MATLAB Answers

0

How to define the sampling time in a system of Simscape Power System Blocks without accuracy lost importants?

Asked by Tony Castillo on 20 Jun 2016
Latest activity Commented on by Torsten Knodt on 5 Jul 2016
Hi my fellows, I'm trying to simulate a subsystem during a time around a whole day (3600*24), but i don't know what is the best way to do it without produce important lost accuracy in the system, because the system collapse the memory of the computer after a time simulated, can you help me with it?, does any one of you has experience in it?. Thanks in advance.

  0 Comments

Sign in to comment.

1 Answer

Answer by Torsten Knodt on 4 Jul 2016
 Accepted Answer

Hi Tony, it's hard to give an advice without knowing your model, simulation settings and the intent of your simulation. Because of the missing information, I just give some generic hints I learned in the past:
  • Rethink your sample times. Eventually some or all can be increased without making your results to bad. If you change the sample time of a discrete SimPowerSystems model, don't forget to adapt the snubber circuits.
  • Deactivate default logging and only activate what is required for your application.
  • Replace the standard logging with explicit Simulink blocks, sending data directly to a file.
  • Use decimation (filters) to reduce the logged data if possible.
Regards Torsten

  2 Comments

Thanks Torten, A question, how do i must adapt the snubber? And the decimation is a problem because when i tried to put it in a value bigger than 1 my signals change too much. Regards Tony
I use the formulas from the Universal Bridge Documentation. In the mean time I also read a bit about different solver methods, interpolation and so on but so far had not the time to test it. It is hard to imagine that you really need all the data taking into account that one would have to view it to make it useful. Could you write something about the model and your application? I bet a somewhat intelligent data reduction (e.g. time filtering around events, result splitting based on events, decimation after low pass or min/max value filtering or other statistics) is the key.

Sign in to comment.