Actual energy generated vs predicted

[DaveDE2]

I've had my system up and running for a bit over two years now and have compared actual energy generated versus what SAM calculations predicted. When I started out, I modeled my situation with SAM (the most accurate predictor that I know of) and thought that I'd be getting output close to that but so far haven't and am wondering if others can respond with what they are actually getting vs SAM or what the installer told you.

I have a 13kW system here in Colorado that should be producing about 19.6MWhrs per year, but in 2017 I only got 17.9, about 10% shy of what I perhaps should have got; this year on track for about the same.

I know this is not much historical data, and predictions are based on historic info but just the same, wondering if anyone else can chime in and say what they got. Wondering if my panels are somehow under producing, etc. Please excuse me if this is a redundant post and direct me to where I can read about it. Thanks.

[J.P.M.]

Over 5 + years, my average daily system output * 365 days has been 9,117 kWh. I've got a pretty good handle on shading loss of ~ 3.5 %. SAM modeling has me at ~ 9,460 kWh/yr for a 5.232 kW system without shading loss.

(9117/(1-.035) = 9,448 kWh/yr.

9,448/9,460 = 0.999.


Something may be different between your SAM input and actual system parameters.

[ButchDeal]

I've had my system up and running for a bit over two years now and have compared actual energy generated versus what SAM calculations predicted. When I started out, I modeled my situation with SAM (the most accurate predictor that I know of) and thought that I'd be getting output close to that but so far haven't and am wondering if others can respond with what they are actually getting vs SAM or what the installer told you.

It isnt a predictor but a model. As a model it is only as good as the model and input. Now SAM is pretty good and likely one of the best of the free models. The best i have found though is aurora solar. Our installs are all within a few percent. We actually use it to validate initial install months and call the installere when things do not line up.

[DaveDE2]

It isnt a predictor but a model. As a model it is only as good as the model and input. Now SAM is pretty good and likely one of the best of the free models. The best i have found though is aurora solar. Our installs are all within a few percent. We actually use it to validate initial install months and call the installere when things do not line up.

Predictor, model, whatever. I created a 3-D model with SAM, measured all my trees, proximity to the panels, etc. The model is accurate. What does aurora do differently? I doubt any installer would take greater care than I did for generating a quote. I'm talking years here not months. Thanks.

[ButchDeal]

Predictor, model, whatever. I created a 3-D model with SAM, measured all my trees, proximity to the panels, etc. The model is accurate. What does aurora do differently? I doubt any installer would take greater care than I did for generating a quote. I'm talking years here not months. Thanks.

Aurora has sub module shadow modeling as well as better 3D simulation and layout. It can input LiDAR which is hard to beat for accuracy, and has really good equipment specifics and calculations.
We have dealings with a few hundred installers and have pretty good accuracy.

The reason it is not a predictor though is the nature of the weather data which is called TYPICAL weather. They choose a typical day from the past measured years to represent each day of the typical year. This is not an average and rarely is there ever a typical year of weathwr. It shuld be pretty accurate on monthly basis assuming weather doesnt change much in a reagoin. However many regions are getting record weather of some kind. Now record is very far from typical.....

[DaveDE2]

I don't see how that's any better than SAM which takes into account years of NOAA weather patterns/data. Lidar scans from an installer from the ground may give some detail but in my case, none better than what I did by measuring my tree height and proximity to the array.

Anybody else want to chime in and say what they're getting vs what they were told, etc? Thanks.

[ButchDeal]

I don't see how that's any better than SAM which takes into account years of NOAA weather patterns/data. Lidar scans from an installer from the ground may give some detail but in my case, none better than what I did by measuring my tree height and proximity to the array.

Anybody else want to chime in and say what they're getting vs what they were told, etc? Thanks.

They ALL (including SAM) use the same typical weather, TMY3 ( typical meterological year version 3)
aurora uses regional plane based LiDAR which is very accurate to a few microns.
Again we have hundreds of installs that were modeled and the only major differences were installers that did something differently than the plan and mistakes.

we also wrote our own pattented model that tries to improve on these models and works at the county / state level. Our system does the modeling for large areas more quickly but again you have the ussue of typical vs reckord weather. Record weather is not going to give typical solar results.

[DaveDE2]

Any other non-trolls here?

[ButchDeal]

Any other non-trolls here?

Wow, you try to help some people.... good luck with your problem!

btw, how many days has your system sat under snow? The models do not account for that well.

[SunEagle]

I believe the problem with any solar estimator is that it uses weather data from the past. Based on what is going on now the weather is different in many ways (colder, wetter, dryer, hotter, more smoke from fires, etc). That difference has to affect any model that uses old historical data.

[DaveDE2]

Wow, you try to help some people.... good luck with your problem!

btw, how many days has your system sat under snow? The models do not account for that.

Of course SAM does.

[DaveDE2]

I believe the problem with any solar estimator is that it uses weather data from the past. Based on what is going on now the weather is different in many ways (colder, wetter, dryer, hotter, more smoke from fires, etc). That difference has to affect any model that uses old historical data.

Agree. That's why I'm asking for input from others as to how their systems have worked over the past few years. Thanks JPM for your input. Any others?

[DaveDE2]

Here's my historical data. The grey line at the top is expected/modeled output. 2016 was a short year, installed mid-summer.

[J.P.M.]

I don't see how that's any better than SAM which takes into account years of NOAA weather patterns/data. Lidar scans from an installer from the ground may give some detail but in my case, none better than what I did by measuring my tree height and proximity to the array.

Anybody else want to chime in and say what they're getting vs what they were told, etc? Thanks.

See my first post to this thread.

I've got lots more data and a fair amount of experience w/SAM. Reread what Butch writes about weather data, and spend some time with the TMY manual to see how it "takes into account years of NOAA patterns data". BTW, most of, or about 90+ % of the TMY data is completely synthetic - not from recorded measurements, except for the 26 original stations. A dirty little secret hiding in plain site with some digging. Doesn't make it bad, as the high degree of correspondence seems to confirm, but it's not data as you may think it to be.

At the end of the day, if you're 10% or so off the model's output - which seems like you have, that's probably within expected variation.

[J.P.M.]

Any other non-trolls here?

I don't think I'm a troll but I'd have been willing to share a lot of information I bet you'd find informative and useful until I read the above post.

Good luck in your future endeavors.