PV watts - is this normal or a bug?

[rsilvers]

Notice how April has more radiation than March and less than May, as expected. But AC Energy is lower in April tan in March.

This is for a southern orientation with a 40 degree tilt.

This does not happen for a Western orientation, so I am not sure it is related to more rain. But maybe it is morning rain?

[sensij]

You can download the hourly output to be sure, but my first guess is that april is warmer than march and hurting the efficiency.

[rsilvers]

But when I change the azimuth to 270, it works as expected:

[sensij]

In the first example, the isolation values are close enough that the effect of temperature can be seen more easily. In the 2nd example, the azimuth affects the isolation and hides the temp effects.

[J.P.M.]

For starters, You do know that those #'s are most likely based on TMY (Typical Meteorological Year) and are not expected conditions for any particular short time period, and also that all the PVWatts #'s are estimates of "long term performance over many years", and even than is further qualified to 10 % to 12 % ?

Best 30 sec. guess at why the model shows April's production being slightly lower than March in spite of insolation being slightly higher has mostly to do with temp., not only ambient temp., but slightly higher panel minus ambient temps. due to a slightly lower daily integrated value of angle of incidence which will increase the daily/monthly time integrated POA irradiance. The general tendancy toward higher panel temps. will lower panel efficiency slightly with an attendant reduction in production. The slightly lower (~ 0.9%) March-April output is about or close to the order of a rounding error anyway.

The TMY data may have slightly more A.M. precip. or more westerly breezes in the afternoon, or less irradiance around noon than other months, or any # of other things. In any event, the differences are quite slight, and such as they may exist in the model seem to have plausible explanations. The precision you may be looking for is probably not warranted, needed or attainable.

Also, per the PVWatts info sheets, given that any month's actual weather and production estimate can be +/- 40 % of the PVWatts long term average estimate and +/- 20% for any year, the diff. of a few %, or less as in this case, is probably not too important in the bigger scheme of things.

Most of this is not rocket science. See the PVWatts info screens and the TMY manual for more information.

[rsilvers]

Ok, I will accept that there is probably no error.

[bcroe]

Notice how April has more radiation than March and less than May, as expected. But AC Energy is lower in April tan in March.
This is for a southern orientation with a 40 degree tilt.
This does not happen for a Western orientation, so I am not sure it is related to more rain. But maybe it is morning rain?

I went through what you are doing a couple years ago. With PVwatts you cannot just make a small input parameter change and expect
a corresponding output response. The reason is, the program adds an abitrary input for the weather, which you cannot see or turn off.
So any small adjustments will be blotted out in the weather changes. Note the big spread they state may happen.

If you try running the hour by hour plots, this will become evident. I find the program useful only for long periods of time where the
weather eventually averages out. There are other simulation programs; I am recording results from real test panels.
good luck, Bruce Roe

[sensij]

I went through what you are doing a couple years ago. With PVwatts you cannot just make a small input parameter change and expect
a corresponding output response. The reason is, the program adds an abitrary input for the weather, which you cannot see or turn off.
So any small adjustments will be blotted out in the weather changes. Note the big spread they state may happen.

This is not correct. The input is not arbitrary, it is supplied by a well defined "typical meteorological year", and is the same from one run of the model to the next if you choose the same location. If you want to see the contents of the TMY file, you can download NREL's "System Advisor Model", which includes the same TMY files used by PVWatts (in an easy to open format).

With SAM (or by other methods), you can eliminate a substantial portion of the weather variance and model the entire year as though it were clear, by building a new TMY file from known clear sky models.

The big spread between actual results and modeled results is because weather truly can vary that much from one year to the next. That is why recording test results from real panels can be misleading over short (monthly or yearly) time frames. The teams that have developed the TMY files review data over much longer periods and use or construct weather patterns that represent something in the middle of the road. *Not* an average, which would be very boring weather, but typical weather that shows some degree of travel away from seasonal norms that you might see in the real world.

If you look at PVWatts hourly output on a clear day, the model variance to actual is substantially less.

[bcroe]

There is of course some issue of what "arbitrary" means. To me it means something of which I was not informed included in the
results. Running results for a clear day works, IF you can first figure out just when/how a clear day is. Good luck finding a clear
year. You wanted results Not slanted by the weather and you DIDN'T GET THEM.

There might be a way to find a "clear year", but I didn't pursue it in depth. Find a place of the same latitude as you, with practically
clear skies (like, a town near DEATH VALLEY); run your simulation there.

When I use test panels, I only take results under a clear sky; in some cases I combined results from a couple consecutive
days that had a few random clouds. good luck, Bruce Roe

[sensij]

There is of course some issue of what "arbitrary" means. To me it means something of which I was not informed included in the
results. Running results for a clear day works, IF you can first figure out just when/how a clear day is. Good luck finding a clear
year. You wanted results Not slanted by the weather and you DIDN'T GET THEM.

In most applications, results that are slanted by typical weather are appropriate for system design. The documentation on the site is very clear about what is or is not included in the model.

[J.P.M.]

There is of course some issue of what "arbitrary" means. To me it means something of which I was not informed included in the
results. Running results for a clear day works, IF you can first figure out just when/how a clear day is. Good luck finding a clear
year. You wanted results Not slanted by the weather and you DIDN'T GET THEM.

There might be a way to find a "clear year", but I didn't pursue it in depth. Find a place of the same latitude as you, with practically
clear skies (like, a town near DEATH VALLEY); run your simulation there.

When I use test panels, I only take results under a clear sky; in some cases I took results from a couple consecutive days that had a few
random clouds. good luck, Bruce Roe

Bruce: You are off the mark.

1.) Reread what Sensij just wrote. It's about as clear as it gets.

2.) Read the TMY manual. It isn't long. It will clear up a lot for you. Honest.

3. While a "clear day " definition is something that has in the past and continues to be a subject of some discussion, the differences in the models are usually of a minor nature - as in, for example, should atmospheric parallax be considered - that type of thing, or differences in the effects of atmospheric constituents, or wavelength bandwidth sizes, or, or, etc.

4.) If you look, clear day models abound. They are usually strung into 365 days of data. I sent you in the direction of several shortly after I joined the forum. As I recall, you were not interested at the time, nor were you successful in locating SAM on the net.

I have a model I did myself in the early '90's. It agrees fairly well with something called the HDKR model. The HDKR model uses calculated estimates for atmospheric constituents such as H2O vapor, mixed gasses, particulate matter, and other stuff. It agrees fairly well with most of the much more rigorous codes.The NREL excel formatted HDKR allows variations in input or uses default values for the atmos. variables. There are lots of models around. Most models output hourly or smaller increments of global horizontal radiation (GHI) which must usually be converted to POA values, especialy for PV or other flat plate estimating work..

Most models also include some estimate of something called a "clearness index (where clearness index == GHI at ground level for a location divided by GHI at the same location above the atmosphere). The seminal work in that regard is considered to be that of Liu and Jordan, and reported in the Journal "Solar Energy", in vol. 4 , 1960.

As a point in fact, most models' output is often a "clearness index", with a "clear day", or clear hour, or any time period for that matter, being one with a "high" "clearness index". My model and the HDKR model seem to give fair, and to my somewhat surprise, consistent agreement with measured and recorded values from the Davis pyranometer located about 4 ft. north of my array. So far, the highest value of hourly clearness index I've recorded is 0.73. "normal" clear day indices around here seem to run somewhere between ~ 0.65 - 0.70 (FWIW, the annual TMY clearness index for the San Diego airport is ~~ 0.59).

The definition and numerical limits used in the various models are somewhat arbitrary. For example, some researchers limit the "clear" day definition to all days whose clearness index is > 90% or so of the max. theoretical value. Other criteria are common.

5.) Point is, and in spite of and regardless of what you and others may think or want, due to atmospheric variations and other things, depending on the level of precision required, there really may be no bulletproof definition of a "clear day" that holds for all climates, locations and situations, but there are enough models, enough experience and enough instruments to measure irradiance to get estimates that have enough accuracy and precision to be useful to estimate insolation both for clear and cloudy days.

[bcroe]

You guys seem to be missing, that we are in agreement. I just explained to the Original Poster, the cause of his results. If you want to help
him find a solution, explain to the OP, not me, the hoops he must jump through to use TMY files, SAMS, NREL excel formatted HDKR, or
whatever is needed. Bruce Roe

[J.P.M.]

You guys seem to be missing, that we are in agreement. I just explained to the Original Poster, the cause of his results. If you want to help
him find a solution, explain to the OP, not me, the hoops he must jump through to use TMY files, SAMS, NREL excel formatted HDKR, or
whatever is needed. Bruce Roe

From my end, think as you wish about agreement. But, FWIW, when it comes to irradiance measurement, how it's done, what's available and it's worth I don't believe you and I are in agreement.

What you "explained" to the OP was, IMO only incomplete and not correct. My effort was to add a bit of correctness and background. Think as you wish, but be prepared to be called on it if someone else sees errors in what you communicate that can lead to misinformation, incorrect actions or poor decisions by those reading what you write.

I respect and appreciate what you've done. I, at least, could probably not do as well. You've been in the trenches and have a lot of hands on, but none of us is as smart as all of us, and in these areas of resource assessment, availability and measurement, IMO, you are not as strong as some others.

In light of your request, consider my response to your posts on this thread as being directed at the OP and others who may have read your posts and acted on what I consider incorrect and incomplete information, not to you, with the usual proviso, take what you want and scrap the rest. The same applies to the OP and others reading my posts. No hard feelings.

[bcroe]

none of us is as smart as all of us

I do know that.

No hard feelings.

I'm having fun here. Bruce Roe

[J.P.M.]

I do know that.



I'm having fun here. Bruce Roe

Fair enough.