How to evaluate health of a new system?

[Alisobob]

solar14.JPG

This is my output for 92656 on a 6.5 KW system.

You can see how weeks 2 and 3 under performed weeks 1 and 4.

Weather made a big difference in the weekly output.

watt.JPG

The only thing that jumps out at me is for these dates, you had sunny conditions and averaged about 15 KwH's per day. For these same dates, I also had sunny conditions, but averaged 28 KwH's per day. A considerable difference.

watts2.JPG

[insaneoctane]

... How are your panels connected?

Honestly, I don't know. I have a single FRONIUSs IG Plus 6.0-1UNI with 24 250W Solarworld panels....I'm not sure the number of "strings" or other details. I don't really have any easy way to test anything on the 2nd story roof, but I have easy access to the inverter. What is the typical hookup for this equipment? Would I benefit from asking any specific questions regarding strings etc?

The most suspicious detail at this point is how low my "peak" output has been since the first week on November; only 1 day with a peak over 4kW (just barely) since Nov 19th. I'll have to keep an eye on the production numbers as it gets sunny...

[sensij]

Honestly, I don't know. I have a single FRONIUSs IG Plus 6.0-1UNI with 24 250W Solarworld panels....I'm not sure the number of "strings" or other details. I don't really have any easy way to test anything on the 2nd story roof, but I have easy access to the inverter. What is the typical hookup for this equipment? Would I benefit from asking any specific questions regarding strings etc?

The most suspicious detail at this point is how low my "peak" output has been since the first week on November; only 1 day with a peak over 4kW (just barely) since Nov 19th. I'll have to keep an eye on the production numbers as it gets sunny...

The typical connection for this system would be two parallel strings of 12 panels in series. That inverter appears to only have a single MPPT, so if all of the panels do not see uniform radiation, it may handle it less efficiently than inverters with separate MPPT for each string.

[Alisobob]

I'll have to keep an eye on the production numbers as it gets sunny...

You had a week classified as "Sunny"... I'd have the installer out asap.

[bcroe]

Honestly, I don't know. I have a single FRONIUSs IG Plus 6.0-1UNI with 24 250W Solarworld panels....I'm not sure the number of "strings" or other details. I don't really have any easy way to test anything on the 2nd story roof, but I have easy access to the inverter. What is the typical hookup for this equipment? Would I benefit from asking any specific questions regarding strings etc?

The most suspicious detail at this point is how low my "peak" output has been since the first week on November; only 1 day with a peak over 4kW (just barely) since Nov 19th. I'll have to keep an eye on the production numbers as it gets sunny...

The arrangement of 2 strings of 12 is likely; 3 strings of 8 is possible. If you have have a combiner
box, you can observe the arrangement, and make measurements there. Your FRONIUS has the facility
to fuse strings and DC disconnect internally. If you see 4 or 6 wires approx 10 gauge coming into it,
those are from your strings. The wires might be labeled, or can be traced to the + and - terminals
visible on the left when the lower door is opened, and the cover removed.

The criteria for clamp on comparisons, is the strings must be identical and in good sun. I'd measure
the + wires here. My FRONIUS 6.0 has bus bars below the + and - inputs, with single heavy wires
going to a remote combiner. Bruce Roe

[SoCalsolar]

here's a source that break down by month

Here's a source that break down by month and by specific equipment, shading etc. Its the CA state calculator.

[sensij]

Here's a source that break down by month and by specific equipment, shading etc. Its the CA state calculator.

Do you mean the CSI EBPP calculator? According to the documentation, it is a repackaged version of PVWatts that allows specific equipment to be selected. The text that matters:

While this calculator relies on industry-standard assumptions, and is driven by NREL’s PV Watts v. 2 calculator, there may be other factors that affect the output of your PV System.

When I set the current version of PVWatts to the CSI default derate (16.3%) and match the inverter effeciency to what is used by CSI, the results come out almost identically. The new version of PVWatts uses a lower default derate (14%), which they have indicated is a better fit to real world data the results have been compared against.

The CSI equipment database is comprehensive, better than what comes packaged by NREL in SAM. Maybe there is a way to update it...

[sensij]

There is a weather station set up not too far from the OP's zip code that reports hourly GHI data. Because the locations are different and the correlation between GHI and energy generation is not particularly straightforward, there are limitations to this approach. The PVOutput.org comparison is probably a better and more user friendly comparison. However, for what its worth, here is a plot of the daily GHI vs the daily energy generated, from 11/1 to today. Over time, if the correlation drifts in ways that can't be explained seasonally or by panel tilt, it could be an indication of a change in performance in either the weather station, or the PV array.

Data.PNG

[J.P.M.]

There is a weather station set up not too far from the OP's zip code that reports hourly GHI data. Because the locations are different and the correlation between GHI and energy generation is not particularly straightforward, there are limitations to this approach. The PVOutput.org comparison is probably a better and more user friendly comparison. However, for what its worth, here is a plot of the daily GHI vs the daily energy generated, from 11/1 to today. Over time, if the correlation drifts, it could be an indication of a change in performance in either the weather station, or the PV array.

[ATTACH=CONFIG]5313[/ATTACH]

Limitations indeed ! IMO, the method is flawed and if followed will probably lead to incorrect conclusions. Over time the correlation will drift and, weather station anomalies aside for now, most of that drift will be quite predictable due to the certain change in the ratio of Global Horizontal Radiation to Plane Of Array irradiance.

Since system output is a function of the Plane Of Array (POA) irradiance, and the POA irradiance is a rather complicated function of not only the Global Horizontal Irradiance (GHI), but also the geometry of the solar position vs. the earth and the array, and a boatload of atmospheric variables, the relative output of the system vs. the global horizontal radiation will most certainly shift because the POA /GHI ratio will shift every day as a function of the declination angle - roughly half the year the ratio will increase and the other half it will decrease. The rate of change will be greatest around the equinoxes and lowest around the solstices.

Some examples: Using the Bird & Hulstorm clear sky model (not TMY data) for zip 92026, 180 az., 20 deg. tilt:

11/01 - POA/GHI ~ 1.32
12/09 - POA/GHI ~ 1.46
12/21 - POA/GHI ~ 1.47

And FWIW:

06/21 - POA/GHI ~ .97

Note the relative change from 11/01 to 12/09 ~= 1.32/1.46 = .90 or about a 10% or so diff.

So, for the dates cited, if the proposed method were to be followed (including the incorrect assumption that POA irradiance changes in the same ratio as GHI as f(date)), aside from problems of site correlation, non clear days and other things, it might lead to the erroneous conclusion that performance actually improved some, maybe by something close to that 10% from 11/01 to 12/09, because of the increased POA irradiance when, in actuality, at least some of the increased output (or a smaller decrease in output if system performance is really going south), might well be due to the increased irradiance in the plane of the array vs. the global horizontal irradiance caused by the constant changes in the declination angle as the earth rotates around the sun.

That 10% or so increase in POA/GHI ratio may more than mask any performance deterioration over the same time period.

In any case, I'd bet it will probably not make the analysis or search for patterns or problems or solutions any easier. Seems to me like it might only muddy the waters more.

Other locations and orientations produce different ratios but the seasonal change in ratios tends to be similar. Given a location, tilt and az., I'll run diff. ratios if requested.

[sensij]

I said it "could be an indication" for reasons that include exactly what you've suggested, not *is* an indication. Prior to your response, I edited the post to more clearly indicate that seasonal and tilt sensitivity (IE, POA irradiance) matter.

Predictable correlation drift is not, in my opinion, a fatal flaw. The most useful way to use the weather station data is probably year over year comparisons... being able to make statements like "this September was sunnier than last September" and supporting it with data can be helpful for people who want to put their system performance in perspective, especially relative to long term estimation tools like those provided by NREL that are not responsive to weather conditions in any given year, and are frequently (and erroneously) held up as a benchmark.

As I said, PVOutput.org comparison may be a better short term indicator of system performance, but knowing that other data sources exist is not, in my opinion, a bad thing.

Although the plot I offered was a simple scatter plot showing the generally good relationship between GHI and generated kWh over 6 weeks, a time series plot of the ratio would probably be a better tool, especially with some SPC type techniques applied to it after correcting for the sun angle as you've suggested.

[J.P.M.]

I said it "could be an indication" for reasons that include exactly what you've suggested, not *is* an indication. Prior to your response, I edited the post to more clearly indicate that seasonal and tilt sensitivity (IE, POA irradiance) matter.

Predictable correlation drift is not, in my opinion, a fatal flaw. The most useful way to use the weather station data is probably year over year comparisons... being able to make statements like "this September was sunnier than last September" and supporting it with data can be helpful for people who want to put their system performance in perspective, especially relative to long term estimation tools like those provided by NREL that are not responsive to weather conditions in any given year, and are frequently (and erroneously) held up as a benchmark.

As I said, PVOutput.org comparison may be a better short term indicator of system performance, but knowing that other data sources exist is not, in my opinion, a bad thing.

Although the plot I offered was a simple scatter plot showing the generally good relationship between GHI and generated kWh over 6 weeks, a time series plot of the ratio would probably be a better tool, especially with some SPC type techniques applied to it after correcting for the sun angle as you've suggested.

I'd agree that a predictable ratio drift is not a fatal flaw, but only if the means to account for it in a quantitative way are used or provided. Without such a way, your method, which did not include such information is of little value and to repeat, may muddy the waters. Therefore, IMO, does not help the situation at hand.

Do as you please, but I'd suggest you learn something more of the solar resource before you lead/point folks off in the wrong direction. IMO, your efforts shown here relating to GHI/POA ratios are counterproductive, and can lead to misleading and inaccurate conclusions relative to the situation. There's already a lot of misinformation out there.

[sensij]

So if I provide too much information, I'm spoon feeding, and if I don't provide enough, I'm misleading. Have you ever been told that you are hard to please? I will follow up later with a post that includes a spreadsheet to estimate the sun angle correction factor you've mentioned, since you find my post useless without it and will offer yours only on an as-request basis. As you well know, there are several models that attempt to approximate the POA irradiance, and although what I use may not be as good as your Bird and Hulstrom model, it does, I think, offer a 1st order correction that starts to make the weather station data more useful.

[J.P.M.]

So if I provide too much information, I'm spoon feeding, and if I don't provide enough, I'm misleading. Have you ever been told that you are hard to please? I will follow up later with a post that includes a spreadsheet to estimate the sun angle correction factor you've mentioned, since you find my post useless without it and will offer yours only on an as-request basis. As you well know, there are several models that attempt to approximate the POA irradiance, and although what I use may not be as good as your Bird and Hulstrom model, it does, I think, offer a 1st order correction that starts to make the weather station data more useful.

I'd bet most of the information you and I are in this latest pissing match about is of little if any concern to most readers. Do as you wish. After this, I'm moving on.

If I'm hard to please at all, I'd like to think part of the cause is that, like my practice of mechanical engineering, I take solar energy very seriously. Otherwise, I'm not that hard to please most of the time, as I've learned to not take myself too seriously.

If you choose to provide more information or less, Quantity is of no matter to me. FWIW, I believe it's a fair criticism that I tend to over explain. But unless you want have you methods called to task for ignorance, more information or less, I'd suggest getting it right, or at least get more informed, or less ignorant about it.

Sometimes, what shows up around here is a good example of a little knowledge being a dangerous thing. Some of my mental spoor included.

Again, do as you wish. Bird & Hulstrom is one of several methods and readily available from NREL. I use a somewhat modified version of it because it seems to be the middle ground of several methods and allows some modification for atmospheric variables.

One last point/comment: Please do not do my thinking for me or imply or accuse me of statements I did not make. First off, I do not find any of your posts useless. The content of your posts for this thread and topic is, IMO, incorrect and can be misleading to many/most folks not familiar with the solar resource. I may think the content of your posts ignorant, but that doesn't make them useless IMO - just wrong. Second, I reviewed this thread several times and I'm unable to find the word useless in anything I wrote. Don't try to put words in my mouth, especially when you don't seem to know the subject I was talking about. Third, I'd respectfully suggest you not confuse serious and honest criticism for ridicule. I offered criticism and reasons for that criticism.

[sensij]

A more general purpose spreadsheet is still forthcoming (grounded in better techniques), but in the course of investigating this further, I thought I'd share another observation.

Using SAM, based on the Sandberg TMY3 file and default loss assumptions, and the array orientation of 171 deg azimuth and 26 deg tilt, it is possible to correlate GHI and kWh generated as I suggested above. As it has been pointed out, and with which I fully agree, the approach is ignorant of the methods used by academics and professionals to approach this problem, and it suffers both from some predictable and from some less predictable errors. However, I believe strongly in the techniques of data comparison to help cancel out some of those errors, and find that sometimes that simplicity in analysis can have value as well.

Anyway, the average daily GHI from the TMY weather file used for November is 3.32 kWh/m2/day. The average daily GHI reported by the weather station identified is 3.39 kWh/m2/day. Based on this, the irradiation used by SAM and the irradiation seen by the weather station may be approximated as equivalent for the purpose of this evaluation.

Using SAM, the average daily kWh output under these irradiance conditions for an array as described is 24.0 kWh. The average daily output reported by the OP is 16.3 kWh. This does seem suggestive that there is some sort of structural loss between the OP's output and the weather station identified. It is hard to say whether that loss takes the form of reduced GHI because of shading or other local environmental factors related to the OP's location, or if it is loss in the actual PV system, also possibly due to environmental factors, along with electrical ones. The decent linear fit and clustering of the data in the correlation suggests that whatever is going on is not something that was introduced in the time period being considered (November). Here is a graphical comparison:

Revised chart.PNG

Investing in a personal weather station might be the best way to understand this more deeply, but perhaps some other troubleshooting steps, as discussed earlier in the thread, might be appropriate first. At any rate, based on this, I am sorry that I suggested earlier that everything is fine. It might very well be performing to the best of its ability, but at a generation rate that is somewhat less that what was sold.

Edit: I didn't have access to a map at the time I was looking at this, and see that Sandberg may not have been the best choice for a TMY3 file. The Camarillo and Van Nuys data may triangulate this location better, especially in elevation, and I will update the results later if it looks helpful.

[J.P.M.]

A more general purpose spreadsheet is still forthcoming (grounded in better techniques), but in the course of investigating this further, I thought I'd share another observation.

Using SAM, based on the Sandberg TMY3 file and default loss assumptions, and the array orientation of 171 deg azimuth and 26 deg tilt, it is possible to correlate GHI and kWh generated as I suggested above. As it has been pointed out, and with which I fully agree, the approach is ignorant of the methods used by academics and professionals to approach this problem, and it suffers both from some predictable and from some less predictable errors. However, I believe strongly in the techniques of data comparison to help cancel out some of those errors, and find that sometimes that simplicity in analysis can have value as well.

Anyway, the average daily GHI from the TMY weather file used for November is 3.32 kWh/m2/day. The average daily GHI reported by the weather station identified is 3.39 kWh/m2/day. Based on this, the irradiation used by SAM and the irradiation seen by the weather station may be approximated as equivalent for the purpose of this evaluation.

Using SAM, the average daily kWh output under these irradiance conditions for an array as described is 24.0 kWh. The average daily output reported by the OP is 16.3 kWh. This does seem suggestive that there is some sort of structural loss between the OP's output and the weather station identified. It is hard to say whether that loss takes the form of reduced GHI because of shading or other local environmental factors related to the OP's location, or if it is loss in the actual PV system, also possibly due to environmental factors, along with electrical ones. The decent linear fit and clustering of the data in the correlation suggests that whatever is going on is not something that was introduced in the time period being considered (November). Here is a graphical comparison:

[ATTACH=CONFIG]5319[/ATTACH]

Investing in a personal weather station might be the best way to understand this more deeply, but perhaps some other troubleshooting steps, as discussed earlier in the thread, might be appropriate first. At any rate, based on this, I am sorry that I suggested earlier that everything is fine. It might very well be performing to the best of its ability, but at a generation rate that is somewhat less that what was sold.

Decent weather stations are about $1K +/- a few bucks and require some regular attention. More for eccentrics like me than 1 time problem analysis. As I suggested on 12/05, it looks like something may be amiss with the OP's system, or shading, or something else, and a call to the installer might be in order as a logical first step. Quality vendors do not usually want unhappy customers.

As for the equivalence of TMY data to actual, unless the weather for the period under consideration in 2014 is the same as the weather for Nov. 2002 - the month TMY 3 used for Sandburg, CA, saying, as you do, that the comparison is valid is the same as saying PVWatts is a predictor of performance over short time periods. As I recall, correct me if I'm wrong, but I believe you and I agree on the futility of that point. It may get close, but that is somewhat a matter of chance - and probably like most moderate climates, not a bad probability for Sandberg, CA, but not a lock.

Furthermore, simply because the numerical average of GHI/day for any 2 days is similar, says nothing about the distribution of irradiance on sunny/pt. cloudy or cloudy days, nor of the intensity of the irradiance. Pt. cloudy skies with, say 50% of the max. daily GHI will produce a different system output than, say, a day with uniform hazy conditions that also results in 50% of the max. daily possible GHI. That is before any adjustments for temp. or wind vector effects on output. Those are a couple of reasons why data comparison as you describe is, IMO only, perhaps invalid for this situation. Reality is likely to be found somewhere between the extremes as few things are all or nothing.

On using DHI data: To eliminate the induced error and added uncertainty caused by using GHI data over periods of more than, say, a few days, I'd respectfully suggest you instead calculate and use POA irradiance and generate your own TMY year using any weather station data of your choosing ( Note that SAM will allow you to do this). Methods to convert GHI to POA are numerous. FWIW, most are somewhat similar in output. The older, less complicated ones (Liu & Jordon for ex.)tend to be less accurate. On the other end, more recent methods are quite complex to program and may not be that much better. More FWIW, I've found the HDKR model seems to give reasonable comparison to my Collected GHI and estimates of POA seem reasonable when comparing system output to estimates. Using the Perez algorithm seems to over predict the POA by a few % or so (or, in effect, under predict system output).