Assistance with Modified Thermal XML Model for Parallel IGBT Modules

PLECS
, , , ,
1

Dear PLECS Support Team

I have developed an AFE converter model that, in reality, uses the 6 Infineon FS450R17OE4 IGBT module. Due to the high current, each switching position consists of six identical IGBT modules connected in parallel.

When I model all six modules individually using their thermal models, the simulation becomes significantly slower. Therefore, I tried to create an equivalent thermal model by modifying the XML file of a single module so that it represents the six parallel modules.

The modifications I made include:

  • Scaling the conduction and switching losses to represent the total losses of the six parallel modules.

  • Adjusting the thermal model by dividing the thermal resistances by six while keeping the thermal time constants unchanged.

However, after making these modifications, PLECS reports the following error:

“Unexpected feedback during calculation of switch losses.”

I have attached:

  • the original XML file,

  • the modified XML file, and

  • a screenshot of the error message.

Could you please help me understand what is causing this error? In particular, I would like to know whether my approach for creating an equivalent thermal model is correct, and whether there are any restrictions on modifying the switching-loss formulas in the XML file.

Thank you very much for your time and assistance.

Kind regards

Milan Van de Walle

FS450R17OE4_switch_modification_to_1.xml (15.6 KB)

FS450R17OE4_switch.xml (15.5 KB)

image
image1190×350 14.9 KB

2

Dear Milan

Your approach is correct in principle, so there is likely a mistake somewhere in the implementation.

Since PLECS 5.0, this kind of paralleling is supported natively through the Number of parallel devices parameter, but only for MOSFETs, diodes, and MOSFET with diode components. For IGBTs, this is less common and still needs to be handled manually.

In PLECS 4.9 and earlier, paralleling had to be implemented in the initialization script. Although the example below is for MOSFETs, the same idea can be adapted for IGBTs:

thermal_parallel_example.plecs (35.6 KB)
C3M0065100K.xml (10.7 KB)

4

Dear,

I’m running into two problems.

Context changes compared to before:

To visualize that fault current (short circuit and earth faults) needs to be interrupted before diode junction temperature gets too high, I added a safety mechanism that disconnects the AFE from both the AC and DC network once diode temperature reaches 150°C.

I have two simulation models:

  1. BASF_1MW_simulated_aangepast.plecs: uses only the I²t value of the diodes, assuming their temperature is already at 150°C when the fault occurs.
  2. BASF_1MW_simulated_aangepast_thermal_eenvoudig_1.plecs: uses the actual thermal model of the diodes to calculate how long it takes to reach 150°C, since in steady-state operation a converter is normally dimensioned so junction temperature stays around 90–100°C max for both diodes and IGBTs.

For the second model, I needed a thermal model for six parallel IGBT modules. I built this using the approach suggested earlier, and put together a test model, parallel_switches_block.plecs, containing an IGBT and diode in parallel (×6). The thermal models needed for this are attached: FS450R17OE4_switch.xml and FS450R17OE4_diode.xml.

Problem 1: Unrealistic ambient temperature required

In the AFE subsystem of (BASF_1MW_simulated_aangepast_thermal_eenvoudig_1.plecs), I have to set the ambient temperature to -275°C before the diode junction temperature drops below 150°C. This makes me suspect my parallel IGBT/diode setup is incorrect — possibly something in the initialization — but I can’t pinpoint what’s wrong.

Problem 2: DC voltage instability in the thermal model

This might resolve itself once Problem 1 is fixed, but I wanted to flag it separately in case it’s a distinct issue. When I compare the DC voltage of the thermal model against the non-thermal model, the results don’t match: in the thermal model, DC voltage keeps dropping and eventually reaches zero — even though no fault is applied yet (it only triggers at t = 1 s, and the simulation doesn’t run that long).

Under normal operation, DC voltage should stabilize at 1100 VDC, as seen in BASF_1MW_simulated_aangepast.plecs, but this doesn’t happen for BASF_1MW_simulated_aangepast_thermal_eenvoudig_1.plecs. I also notice significantly more current being drawn from the AC grid, yet the DC voltage still collapses to zero.

Could you help resolve these two problems? Thanks!

Kind regards
Milan

parallel_switches_block.plecs (68.3 KB)

BASF_1MW _simulated_aangepast_thermal_eenvoudig_1.plecs (701.9 KB)

BASF_1MW _simulated_aangepast.plecs (363.1 KB)

FS450R17OE4_diode.xml (9.7 KB)

FS450R17OE4_switch.xml (15.5 KB)