Best Practices for using pan file specs vs. model parameters for module model

  • operaman95
  • Topic Author
16 Aug 2023 20:35 #12425 by operaman95

Thank you for maintaining this excellent forum.

I've recently been trying to evaluate new First Solar Series 6 bifacial modules using preliminary pan files and several PySAM modules. I've been using the IEC61853 module model to do this. I've been taking the following steps:

1. Parse the pan file and isolate key module parameters (Voc, Isc, RSerie, RShunt, etc.)
2. Run the module parameters through the `SixParSolve` module to get the parameters I need to run the `pvlib` single diode model, which I understand is the SAM/CEC model.
3. Run the SAM/CEC model across IEC61853 test conditions.
4. Use the results from 3 to run the `IEC61853par` module to derive Rshunt and Rseries fitting coefficients (D1-D3, C1-C3)
5. Finally run a full energy simulation with `pvsamv1` and the IEC61853 module model with a mishmash of parameters that partially come from the pan file and partially from various steps in the modeling process.

As I've done this, I've learned that my results only make sense when I use the reverse saturation current value (RDiode in panfile, Io in single diode equation nomenclature) from the `SixParSolve` module and NOT the current from the PAN file. These two values end up being quite different (1e-11 vs. 2.8e-12). Furthermore, after running the `Iec61853par` module, I get an Egref (band gap) of 1.0958, fairly different from the 1.45 I understand to be the empirically accepted band gap for CdTe tech. 

Could you please give me some guidance on if the procedure (above) I've come up with for running the `IEC61853` module model starting with a PAN file makes sense? 



Please Log in or Create an account to join the conversation.

  • jdnewmil
12 Jul 2024 20:53 #13309 by jdnewmil
When you apply linear regression with a third-order polynomial, the second-order regression coefficient you obtain will be different than the second-order coefficient you get when regressing a second-order polynomial. In the latter case the second order term has to approximate the behaviors that were separate in the former case, so the coefficients are not comparable.

PVsyst uses a single-diode model that is augmented in several ways that may make coefficients obtained for it rather different than coefficients you might obtain with some other "single-diode model". Some key differences are the Rsh(G) not being a constant (, and the recombination model ( used with thin-film modules. So the first solution is not to get hung up on individual coefficients, and if you must in order to be compatible with PVsyst then you have to assume the right form for the model when extracting coefficients from raw data. I am not sure how well the PVsyst model has been reverse-engineered to support this... you may get best results using PVsyst itself to extract coefficients from test data interactively.

Please Log in or Create an account to join the conversation.

Moderators: pgilman
Powered by Kunena Forum