Generate large AI training dataset from PLECS (parameter sweep + peak current extraction)

General
1

Hi Plexim team / all,

I’m trying to develop a workflow to generate simulated data from a DAB converter under triple-phase-shift modulation.

What is the recommended way in PLECS Standalone to run a large parameter sweep (tens or hundreds of thousands of runs) and return signals to a script efficiently?

For each simulation, I only need the steady-state peak current (e.g., the last few switching periods).

Thanks in advance!

2

Hi @maaqibsh

I would strongly recommend taking a look at the demo model “Buck Converter with Parameter Sweep.” This example highlights several useful aspects that are directly applicable to your use case:

  • External scripting via XML-RPC or JSON-RPC. The demo uses Python, but any language that supports XML-RPC or JSON-RPC can be used.
  • Parallel simulations. You can run up to number_of_CPU_cores simulations in parallel, which is essential for large parameter sweeps.
  • Efficient data extraction. The model demonstrates how to extract only specific scalar results instead of logging full waveforms, which significantly reduces overhead.

If you have any further questions, feel free to let us know.

Best regards,
Reto