Snow loss modeling

[jackattack]

I would be very interested to finally see snow losses calculated in PVsyst, either through the link with Meteonorm (the TMY2 dataset contains snow depth info I believe) or allowing import of snow data by users, and having the software calculate losses. This functionality exists in System Advisor Model but it would be good to see it here to make sure PVsyst remains the premier modeling software. BEW and others have provided formulas for calculating losses that could be included in PVsyst. As we work more in northern climates, this becomes a liability for us.

I would be open to hearing how you or other members calculate for snow losses currently.

Thanks!

[unilhexio]

In the case where applies, I introduce it as seasonally soiling losses

[jackattack]

I agree that that is the correct place to add in snow losses, however I think PVsyst is in a position to help calculate these losses using established formulas added into the software.

[André Mermoud]

I agree that a model would be very useful.

However this would require several input informations, namely in the meteo file, but not only.

The effect of snow is not only the falling snow, but also the remaining snow on the PV modules (how much time will is stay? Partially or completely? Effect of the temperature, insolation, plane tilt, module's base geometry, etc.

We will think about implementing a snow coverage ratio defined in hourly values, to be specified by the user. The difficulty will be to evaluate the electrical mismatch effect in case of partial covering.

[jackattack]

Thank you Andre for your response. I agree it is a very complicated formula and may vary in accuracy. I think partial coverage is the biggest issue as well. The paper by BEW and DNV a few years ago provided what I think is the best existing formula, at least as far as I have seen. I believe NREL has issued some papers as well. I wuld be happy to send you those papers if you're not familiar with them.

It would require specific inputs from the user but could be feasible.

[MarcoMalagnino]

Dear all,

The exact forecast of the snow amount, in term of losses and periods, is quite impossible. There are some possibilities linked with meteo data which indicate also snow coverage, but as Mr. Mermoud has written, there are several different factor which influence the snow coverage effect: snow type, ambient temperature before, during and after snowing, module tilt angle, wind, humidity, irradiation, etc.

I think that the evaluation of the expected losses due to snow coverage is responsibility of the user, as well as for normal soiling (also difficult to appreciate as the snow).

For snow losses, probably could be a good approximation to define some days per month with the expected snow coverage of the month (a sort of unavailability of the system) and concentrate the expected monthly snow losses in the selected days of unavailability, without taking into account the partial covering effect. The hourly or daily behavior would be probably not realistic but the monthly values could be good enough to evaluate possible influences on the design of the pv-plant (narrow row distance, higher tilt angle, etc.).

However I think that the snow losses, as well as the soiling losses, should be taken into account by PVsyst before the other losses (Horizon, Shading and IAM losses).

Best regards

[jackattack]

Marco,

I have to disagree - it is not impossible - several studies have shown a statistical method to account for all of the factors that you mention, notably the BEW/DNV study, and NREL studies. It is not perfect, but neither is the TMY. It's a best guess. And, System Advisor Model already provides snow loss calculations using the TMY2 dataset which includes snow depths.

I agree that is the responsibility of the user to provide the data to run the snow losses, but I would simply ask PVsyst to embed one of the statistical models within their program to make things simpler and easier. The same is true for soiling losses.

Obviously snow depths can vary considerably from year to year - from my hometown in Minneapolis, Minnesota, we see snowfall vary as much as +/-30% year to year. But, being able to include a typical estimate is still better than nothing!

[Humayun]

Snow soiling has tons of variables. I believe all three mainstream snow soiling models (Townsend, Marrion, and Andrew's) has very general approach and they do not consider the following:

- Falling and accumulated Snow behavior on fixed tilt ground mount system

- Falling and accumulated snow behavior for single or multiple axis trackers on ground mount systems

- Falling and accumulated snow behavior for pitched roofs with flush mounted systems

-Falling and accumulated snow behavior for flat roofs with certain degree tilted arrays. Flat roof building brings more challenge because snow has no where to go and if it is ever windy snow drifting keeps snow moving from one space to another

In Canada and northern states realistic snow soiling in Jan, Feb, and March can be anywhere from 90% to 100%. Challenge is to estimate other wintry months. Anybody know of any other snow models then what I listed above?

Cheers,

Humayun