Incoherence in loss calculation

Question

Even with a simple model of inverter, the results proposed are not coherent in comparison with the theory... I don't understand what happen because I took the Buck model in the Demo and I adapt to my inverter. Could you help me please and explain what's wrong with my simulation?

Comments

You will really need to post a model for anyone to help you with your question.

Answer 1

Sorry, can you open it?

Comments

In this model the thermal results seem reasonable and the junction temperatures remain fairly low in steady state and do not diverge.  The issue here is the efficiency calculation will be off since the current delivered from the source during a switching cycle is both positive and negative, leading to a low average power (~20W) but high peak power (~ +/-5000W).

I also am not sure the modulation scheme here is achieving what you would like, but after tweaking the scheme to have 10A DC average current flow the results still seem reasonable.  

Also note in your efficiency calculation you have a scaling factor of 6 to account for the device loading, but there are only 4 devices here, but with a different scheme the devices will be loaded differently and this assumption will no longer be valid.

Regarding your other model, I agree with Kris that your controls do not look correct (for example the grid current is not sinusoidal) and that must be resolved before investigating the thermal performance.

---

Thanks a lot. Can I unextrapolate the curves to obtain the real curves? Presently switching are perfect whereas I should observe a commutation time?

---

What curves are you referring to?  Note that the device models in PLECS are behavioral representations of the thermal device characteristics but not reflect the device switching dynamics.

---

Thank you for your help but I still don't understand how to fix the block periodic average and periodic impulse average. Besides, even if my step is 0.5e-8, the simulations results show a point every 0.5e-6. I can't therefore deduce my switching losses... and th result s similar at 16kHz and 50kHz... Thanks for your help

---

Furthermore, when I replace MOSFET by IGBT, it displays Thermal states derivatives are infinite or NaN... how can I fix this problem? I've tried to change the frequency and the Rth...

---

The model you shared has several fundamental issues that must be resolved before you will see reasonable results.  After making the following changes the thermal transients make sense to me.

1. You should be using a variable step solver instead of a fixed step solver.
2. Part of this is that your controller is not setup appropriately.  Did you look at your Vd and Vq values going into the modulator?  They are not stable and change every time step.  This results in an effective switching frequency in the MHz range (depends on the fixed time step, but explains why the switching frequency has no impact on losses). One fix that seemed reasonable and worked for me was to use a variable step solver and place a 1/VDC scaling factor in the feedback path, but ultimately you must design this system.  For example I think your carrier limits should be +/-1 instead of +/-0.5.
3. The averaging intervals for all thermal loss calculations should be the same and correspond to the switching period. Your averaging intervals do not agree.

I do not understand your issue with the MOSFETs vs. IGBTs.  Some of the above changes may help you resolve the issue.

 

In the future, if your model uses thermal models create a *.zip archive structured like the one I have attached with the models included.  PLECS searches the <model_name>_plecs folder for the thermal descriptions and allows other people to run the model with less effort.

---

Thank you for your help.

There is still some incomprehension... my output seems valid but I don't understand my IGBT conduction losses... why are they so small?

In y folder, I've joined the IGBT. I'm using the IKW25N120H3_IGBT

Thank you for your advice and I hope it will be easier for you