Adding New Single Diode Model For CdTe

[crimmj]

Would like to add a new diode model(s) appropriate to non c-Si modules. There are some additional parameters in the model(s). Any thoughts on the correct structured way to do this? Would this function or functions be added to pvsystem.py? Do modules need to be a class if they have different appropriate performance models?

[wholmgren]

Sounds useful to me. The existing functions that use module parameters will work with anything that supports dictionary-like lookups. Most of the time, people will be passing in Series objects from the SAM DataFrames, but plain dictionaries should work too. I would say that we should keep things simple and use a similar interface for new functions.

[cwhanse]

This request essentially asks to implement the PVsyst model for a-Si, which includes a new term to account for recombination loss in the intrinsic layer (see http://files.pvsyst.com/help/thinfilm_recombinationloss.htm

PVsyst provides a default value for the single new parameter (di^2/mu teff = 1.4V).

Implementing this new model depends on having calcparams_PVsyst see #470 and will require a new function to solve the diode equation with the recombination current term.

[mikofski]

I have this model already, I can try to add this

[cwhanse]

@mikofski happy to have a volunteer! Are you thinking to implement the PVsyst model (addressing both #470 and #163) or only the variation of singlediode to solve the current-voltage equation? If you need help with #470 I can do the calcparams_pvsyst component.

[mikofski]

If the recombination loss current is already included in the MATLAB version, then I guess we should just port that over here. @cwhanse are you planning to do that or are you looking for volunteers?

[cwhanse]

I have some MATLAB code which isn't in PVLib. I didn't find an elegant solution, it's brute force getting the current at each voltage.

[mikofski]

Okay let's collaborate then. I use the same explicit approach for PVSyst as I do for any SDM. Then either interpolate or Newton method or bisection for specific points or the max p. I'll try to put something up this week. I think this counts as work don't you? On a related note, I added an issue on SAM SSC on the recombination current for thin film. I think it should be added to the CEC and DeSoto models, the J. Merten, Mermoud, and Lejeune seem to indicate that an error would be present with thin film for any SDM near Voc missing this term. Do you have an opinion? Maybe we can collaborate with FirstSolar.

[cwhanse]

I think I need to be clear on the plan here. This issue asks for a new function (counterpart to pvsystem.singlediode that solves

I = IL - IL x mu/[Vbi - (V+I x Rs)] - I0 x [exp((V+I x Rs)/(nNsVth))-1] - (V + I x Rs)/Rsh

returning the IV curve. The second term on the right is the recombination current.

#470 asks for calcparams_PVsyst which is related, but basically independent, because the parameters unique to the recombination term (mu, Vbi) are constants.

We can work together to implement singlediode_recomb or whatever name we settle on. I'll email my Matlab function to you.

[mikofski]

Yes, I'm clear. What I was asking about was that since the only difference between the existing single diode model (SDM) and a model that included the thin film recombination loss current is the term:

I_rec = I_L * d2_mu_tau / (V_bi - V_diode)

then:

• does it really make sense to create a separate sdm_thin_film method that is nearly identical to the existing SDM method?
• would it be possible to just include I_rec = I_L * d2_mu_tau / (V_bi - V_diode) into the existing pvsystem.singlediode and just turn it on as needed?

As I said previously, the solution for either is exactly the same: specify a bunch of diode voltages, solve for cell current then back calculate the cell voltage. For specific voltages and max power, then you can either interpolate, use Newton, or use a bisection search method. (Maybe Lambert-W also works but I'm not as familiar with that formulation?)

I understand that calcparams_pvsyst merely generates the 5 parameters that SDM uses since PVSyst has different auxiliary equations, so I understand that's the focus of #470.

So, as I asked above, for #163 we could either implement a new SDM method for thin film with recombination losses or just combine it with the existing SDM.

Regarding:

I think it should be added to the CEC and DeSoto models, the J. Merten, Mermoud, and Lejeune seem to indicate that an error would be present with thin film for any SDM near Voc missing this term. Do you have an opinion? Maybe we can collaborate with FirstSolar.

Sorry to be so vague with that comment. What I meant was the question I asked above: shouldn't we just include the recombination loss current into the existing pvsystem.singlediode such that it would be available to any calcparam_*() method, not just PVSyst? I was thinking that by collaborating with FirstSolar we could determine whether or not the existing SDM without the recombination loss current exhibits the errors that J. Merten, Mermoud, and Lejeune observed.

[adriesse]

I think it makes sense to incorporate this feature into the SDM functions. The various models can then use the feature as needed. Maybe it even opens up some unexplored territory.

[cwhanse]

I'm tracking with you now @mikofski If we use the diode voltage method to solve the IV curve this new term is not a challenge. But singlediode and its helpers, i_from_v and v_from_i are using the LambertW method which doesn't extend when the recombination current is added to the single diode equation.

We had a long conversation about changing to the diode voltage approach #409 which hasn't been merged. We need to resolve that before moving ahead with this issue.

Since it has been a while, @mikofski can you remind us what issues remain with #409?

[mikofski]

I've started trying to update #409, there's a list of TODO[s] that @wholmgren left for me to check off like fixing docstrings, better names (newton vs. "fast", etc.) and moving vendorized scipy.optimized.newton to _tools since it won't be merged until November (if ever ☹️ )

[mikofski]

TODO: see this comment from #409 to add Irecomb to bishop88

Yes. Set d2mutau=0.0 and Vbi=0.9, user can change the values with kwargs if wanted.