PVWatts doesn't model snow shading

[nerdralph]

I've been looking at rooftop PV monitoring data for systems here in the GWN, and and noticed most of them produce less than expected energy during the winter months. After doing some more research, I found a paper that did a multi-year experiment in Keshena, WI. It found that failure to model snow shading in PVWatts over-estimates production by 3-4%.

https://docs.lib.purdue.edu/cgi/viewcontent.cgi?article=1003&context=tlipubs

[J.P.M.]

As a practical reality, the decrease in system energy production due to snow loading/shading is not something that can be modeled with much hope of success.
It's a bit similar to earthquake prediction and has some of the attributes of chaotic behavior, even more so than the variability of irradiance or precipitation.

[bcroe]

The amount of error caused by snow will vary with many things. Perhaps the
most significant is how the arrays are configured for winter. Same as summer,
not good. There is the weather, which can vary so much. First there is the
kind of snow fall, then how much and soon sun may be able to clear it. Ice
cannot be cleared until the right combination of sun and warmer weather
comes along.

Arrays here are specifically configured to reject snow. With me out there clearing
the morning after, snow has virtually no effort on my system. My neighbor has
panels virtually on the ground, and very close together. It is virtually impossible
to get in there and clear them. Bruce Roe

[J.P.M.]

The amount of error caused by snow will vary with many things. Perhaps the
most significant is how the arrays are configured for winter. Same as summer,
not good. There is the weather, which can vary so much. First there is the
kind of snow fall, then how much and soon sun may be able to clear it. Ice
cannot be cleared until the right combination of sun and warmer weather
comes along.

Arrays here are specifically configured to reject snow. With me out there clearing
the morning after, snow has virtually no effort on my system. My neighbor has
panels virtually on the ground, and very close together. It is virtually impossible
to get in there and clear them. Bruce Roe

Like I wrote, hard to model.

[nerdralph]

pvlib added snow shading in v0.7.2. I'm going to try it out.

https://github.com/pvlib

[Preaky]

I used a free trial of a software called pylon. It let's you map trees by type (palm, spreading, other types) etc, poles or buildings nearby.

Super easy to use shading tool.

I will be installing system with with building and trees of various types around my house - I can scroll through the hours of the day and on different days of the year to see where shade is happening. Also you can oversize your trees, or remove obstructions.

Then it takes your shading inputs + historical irradiance and give you shading results - total, by time of day, day of the year.

Very cool simulations. I think manually inputting obstructions with something like Pylon is better than LIDAR used by Aurora and OpenSolar.

How accurate it is, time will tell. For example, using the software I estimate I will have >50% shading in January. <5% shading in the summer months. And I will have approx 20% loss to shading over a calendar year. Ultimately, how accurate it is - time will tell for me. Still seems like the best tool for estimating shading out there. And free fully functional trial, if you have a few hours to tinker, definitely worth it.

[nerdralph]

I've been using OpenSolar for my PV projects. It supports manual shading analysis, but I haven't played with it much at all. The solution to trees is my Stihl!