That's why I was thinking about how to use insolation data (which I don't have anyway) to correct for this and come up with an estimate of the accuracy of the estimate.
Most Popular Topics
Collapse
Pre-installation Production Estimates vs. Reality
Collapse
X
-
CS6P-260P/SE3000 - http://tiny.cc/ed5ozx -
This is the closest weather station to Minneapolis I've found reporting irradiance data. With free registration, you can download data back to 2007. If you can find a meaningful way to use this before I can, please share it! I am working on it for fun, but am not confident that I can account for enough of the error sources and other factors to make any meaningful statements about array performance. I am fortunate that near me, more stations are around that report irradiance, so some averaging can smooth things out.16x TenK 410W modules + 14x TenK 500W invertersComment
-
Thanks! Life got in the way...
Probably true.
I was trying to come up with a way to compare the estimate to real-world production data to determine how accurate the estimate was. So the production estimate is an input, not an output.
This is the part I'm not understanding. If I'm taking the estimate as an input and trying to compare it to real-world data, why do I need to know the details of how the estimate was produced? It either is close to the actual production or it isn't.
Of course, I realize that with weather being weather it's not reasonable to expect five months' of production data to be within 20% of the original estimate. That's why I was thinking about how to use insolation data (which I don't have anyway) to correct for this and come up with an estimate of the accuracy of the estimate.
Before signing the contract, I actually did plug in the installer's parameters and shading factors into PVWatts to verify that I got the same results. And they were within a percent or so.
But all that tells me is that the two different tools agree with each other, not whether they agree with reality. For example, if inputs like the efficiency of the inverters were over/underestimated, the tools would agree with each other but not match reality.
After only five months of non-average weather, at this point the estimate could have been wrong by 20% in either direction and I wouldn't know. I could of course just wait five years and average my annual production over that time to get a pretty solid answer as to whether the production estimate was accurate. But I'm not that patient.
OK. Take a deep breath.
To estimate system efficiency you will need two things: system output and system input. Divide the first by the second and stick a fork in it. Do that for periods of a minute, hour, day, month, or longer.
Reality is often a matter of perception. If you want real world production data, you will need to define at some point how much your system is producing. This amounts to saying your system monitor, or other means of measuring your output is reality.
You will also need to define what your system input is - how much solar radiation your array is getting. This may be the tricky part.
Here's why: The best way is to get yourself a pyranometer. So far, so good. The fun starts when you adjust the horizontal solar radiation the pyranometer measures to something called "Plane Of Array" (P.O.A.) radiation intensity on an hourly or more often basis, multiply that by your array area, and define that as your input. You can use remote sites and other pyranometer and weather data, but the P.O.A. calculation is sill necessary, again on probably an hourly basis. As Sensij suggests, see Duffie and Beckman for what's involved, and be patient. I'd do that to get a grasp of what's involved before I used available software, which is often of dubious use and accuracy. Some of the NREL stuff is pretty good however.
Summing those results for one period of time (say, one measurement at the midpoint of an hour) will give you something called "instantaneous efficiency" or "hourly efficiency". Summing hourly data and results over longer periods will give efficiencies over daily, monthly or longer periods.
Bottom line : If you are looking for a way to compare your measured system output against the output of an estimating program like PVWatts and then use some "adjustment factor" or "magic number" to get the two to match up or somehow correlate in a recognize and reliable fashion to see how your system is doing against what you were told it would do without modifying the solar radiation input to the estimating program - I do not know of any such method and I doubt such a procedure exists. But I could be wrong. I seriously doubt it on this one, but you never know.
Here's one reason why: To check on the accuracy of any estimating program against what you measure and define as reality, any such check must include the same solar radiation input as the pyranometer output adjusted to P.O.A. you used both in magnitude and time distribution. Some of the more sophisticated estimating tools will allow that. SAM does, but it's sort of a PITA.
If the estimating program solar P.O.A. input and what you define as your P.O.A. input are identical, the difference between what your output monitor reports and what the estimating program pukes out is a measure of the validity of the program output - i.e., its accuracy.
Q.E.D.
Note: If you do hourly data, don't be surprised if the individual hours don't match as well as the summations.
To my experience, a few of the things about solar energy applications germain to this discussion that most people, maybe including you, are not aware of may include:
- Solar simulation and estimating programs like PVWatts, SAM, TRYNSYS and others do NOT use average values of solar radiation. Mostly, they use what may be called "representative" values for solar radiation and other weather variables. Most of it is measured, but may have significant gaps. For example, the TMY data that PVWatts and SAM usually use is generated by picking the most "representative" month ( say, Jan., 2004 and Feb., 1997, etc.) of the last 15 20 or 30 years (how many years to compare depends on some other things) for a particular location, using some selection criteria weighed in favor of solar radiation, and stringing the "most representative" 12 months together to form a "typical" year. See the PVWatts help screens or better, the TMY manual for details.
- Converting TMY, or other data (such as from pyranometers at adjacent or remote sites) into something useful and indeed necessary for solar simulation or estimating schemes (P.O.A. radiation) is by no means a straight forward, well defined or agreed upon process, but it is essential. Fortunately, most simulation programs do that task fairly well. Most knowledgeable folks in the business know that some algorithms are better than others for different conditions. Usually, various methods produce results that are somewhat similar to one another. For example, one well known and respected method - the Perez method - gives similar, but different results than something called the HDKR method. Those 2 usually agree within a few %, with Perez often giving higher rad. est. The few % may seem like small potatoes but if you're trying to figure if system efficiency is, say, 16% or 18%, and 2 respected methods mentioned here are, say, 5 % apart, things get less certain quickly. Example: Do a SAM run and change the input from HDKR to Perez and see what happens to the yearly hourly output total and scatter. BTW: the waters get muddier still - most commonly used pyranometers on the Davis weather stations are good to ~ +/- 5% or so, and then, only if kept in calibration and checked every year or two. Hopefully, the precision is better than that. Even the high $$ Eppley instruments (~$5-$10K+) are probably no better than 2-3% or so. FWIW, my pyranometer, located about 4 ft. north of my array seems to be pretty close to what's predicted by clear sky estimating algorithms.
Take what you want/need of the above, scrap the rest.Comment
-
dec 2014.JPG
dec 2014.2.JPG
solar2.JPG
Forecast was 688 KWh's
Actual was 581 KWh's.
Weather was bad for several days....
It plays a huge role, in hitting your expectations....
Lets wait and see if Jan. can pick up the slack!!!Comment
-
It is sad that immediately after a post in which J.P.M. has concisely shared so much information about how PV estimating works and ways one might validly approach it, someone could still use the word "forecast" in context with PVWatts modeled output. PVWatts isn't a predictive model. Keeping month by month score is fun, but foolish, but probably harmless except in that it perpetuates misunderstanding of the tools NREL has made available to us.CS6P-260P/SE3000 - http://tiny.cc/ed5ozxComment
-
to calculate or predict (some future event or condition) usually as a result of study and analysis of available pertinent data; especially : to predict (weather conditions) on the basis of correlated meteorological observations; to indicate as likely to occur… See the full definition
The 3 things that directly effect solar output are sun intensity, cloud cover, and ambient temps.
If that isnt a forecast, I dont know what is.Comment
-
http://www.merriam-webster.com/dictionary/forecast
The 3 things that directly effect solar output are sun intensity, cloud cover, and ambient temps.
If that isnt a forecast, I dont know what is.[SIGPIC][/SIGPIC]Comment
-
http://www.merriam-webster.com/dictionary/forecast
The 3 things that directly effect solar output are sun intensity, cloud cover, and ambient temps.
If that isnt a forecast, I dont know what is.
fore·cast
verb \-ˌkast; fȯr-ˈkast\
to say that (something) will happen in the future : to predict (something, such as weather) after looking at the information that is available
Originally posted by J.P.M.- Solar simulation and estimating programs like PVWatts, SAM, TRYNSYS and others do NOT use average values of solar radiation. Mostly, they use what may be called "representative" values for solar radiation and other weather variables. Most of it is measured, but may have significant gaps. For example, the TMY data that PVWatts and SAM usually use is generated by picking the most "representative" month ( say, Jan., 2004 and Feb., 1997, etc.) of the last 15 20 or 30 years (how many years to compare depends on some other things) for a particular location, using some selection criteria weighed in favor of solar radiation, and stringing the "most representative" 12 months together to form a "typical" year. See the PVWatts help screens or better, the TMY manual for details.
In general, you can expect the system's total electrical output for a given month of a particular year to vary by as much as ±30% from the long-term typical value. Similarly, the total annual output for a particular year may vary from the long-term typical value by as much as ±10%.CS6P-260P/SE3000 - http://tiny.cc/ed5ozxComment
Comment