Selling home, solar inspection issue with VOC measurement.

[max2k]

Read the first post again. Ambient temp was 75-80 deg (25 deg C) at the time of the measurement, on the way to a day in which temp would be well over 100. The NOCT rating of the panel is made at 20 deg C ambient, yielding a cell temp of ~46 deg C, so if rooftop temp was equal to the 25 deg C ambient temp, you would already expect the cell temp to be over 50 deg C. Rooftop temp is typically warmer than ambient, so even if you don't get all the way to the 66 deg C you are estimating, *most* of the reduction in Voc observed can be explained by temperature.

too many assumptions- no way cells were near 46C at 8:00 AM in the morning and it doesn't matter how hot the coming day would be, more important how hot the previous night was. With 25C ambient 35C cell may be but that would give drop of only 10x0.004x30 = 1.2V per panel. I can easily see low irradiance to be prevailing factor as my own array produces < 5% power around 8:00 AM from its middady peak and 20% irradiance level already gives 5-7V drop. Of course roof orientation and local weather patterns play big role here.

[bcroe]

when load is short circuited the energy it dissipates is 0 so it must go elsewhere. I was referring to this specific case only.

Well you may be right for the shorted situation. My thought was that there was no voltage drop
for current passing through the diode. Now I suppose I have to run the experiment to see
what happens. Bruce Roe

[sensij]

too many assumptions- no way cells were near 46C at 8:00 AM in the morning and it doesn't matter how hot the coming day would be, more important how hot the previous night was. With 25C ambient 35C cell may be but that would give drop of only 10x0.004x30 = 1.2V per panel. I can easily see low irradiance to be prevailing factor as my own array produces < 5% power around 8:00 AM from its middady peak and 20% irradiance level already gives 5-7V drop. Of course roof orientation and local weather patterns play big role here.

Look at his PVoutput data. At 8 am, the system is already producing over 30% of peak power. As J.P.M. estimated, POA irradiance is probably over 300 W/m2 at that point, so you are probably looking at only 1-2 V change in Voc due to irradiance effects. I can spend more time tuning the model for this specific installation just to prove the point, but I think you are barking up the wrong tree here, and underestimating the thermal effect.

What do you think the thermal time constant of the PV cell is?

[J.P.M.]

I think it is important to stress that temperature and irradiance cannot be so easily decoupled in the real world as they are in the analysis I provided. Even within the CEC model, if I attempt to keep environmental conditions constant and change only effective irradiance by changing the panel tilt (for example), the impact of cell temperature on Voc overwhelms the impact of irradiance, and except at the extremes, the Voc is largely determined by the simple temperature coefficient. So, no, I don't think it changes any of the advice to the OP, but I do appreciate that the truly accurate response is more nuanced than early thread replies suggested.

(In other words, the information provided by suncalc.org doesn't do anything to help the OP reconcile the electrician's measurements with what reasonable expectations of the system Voc should be)

I've got something to add to the connection between panel or array temperatures as measured, and theory as it currently exists. FWIW, I've used theory, a lot of it from the very same sources Sensij quoted from to develop a methodology to correlate array temperatures and array voltages. I needed to develop that to be able to estimate array fouling and investigate how array fouling changes as f(time, weather conditions). I then tested the method by measuring each of my array's 16 panel's temps. in 4 random places on each panel over a 6 to 8 minute time period both before and after the time of minimum incidence angle (16 X 4 X2 = 128 temp. measurements per day's trial, all averaged together and used as that day's average measured temp.) Then, at the time of min. incidence angle for each of those days, and between measurements, I recorded string voltages and current at the inverter input, string power D.C. inputs to the inverter, and total A.C. output from the inverter for well over 100 clear days. I winnowed those 100 + measurement days down to 60 days of such measurements split about evenly between winter and the following summer, with all days in each season being consecutive clear days. In that sense, beyond the start and stop of each season and the goal of ~ 30 measuring days in each season, I didn't cherry pick the data. I also used on site weather data, including GHI, taken continuously in one minute intervals from a Davis instruments Pro II Plus weather station located about 4 ft. north of the array at the same elevation as the highest array point.

A couple of results: Array or sting current correlates very well with POA irradiance. After confirming the d(current)/d(irradiance) ~ = +.0035 A/deg. as per the spec sheet (but that was hard to measure and I can't confirm it beyond saying it isn't much), I got a current change per irradiance change based on calculated POA irradiance of ~ 164-169 W/A-m^2), vs. a spec sheet value of (1,000 W/m^2)/5.98 Amp = 167.2 W/(A-m^2). I'd suggest that if it can be compared to and calibrated against a pyranometer, an array's output can be used as it's own pyranometer.

Bottom line: Theory as I understand it matches most all the data I've taken, and vice versa to a better degree than I had expected. I believe I confirmed the spec sheet claims reasonably well, including not only that current does not vary too much with temperature but directly with POA irradiance, but also that the second derivative of voltage with respect to temperature is close to zero, at least for the array temps. I measured (33.5 C- 67.1 C.). I did some farting around with the numbers, just for the hell of it, and SWAGed 2d deriv. value of volt/temp. of ~ -2.4 EE-4, but that's probably no more than B.S. that's not defensible given the instruments I'm using, but I had fun doing it and refreshed some math along the way.

One semi detailed example of what I got from all these measurements: String Voltage increment per degree of array temp. difference worked out to have an average value of -1.477 v/deg. C ( std. dev. = 0.123V/deg. C, N=60). The spec sheet per deg. C Voltage drop is listed as -0.1766V/deg. C. per panel. --->>> 8 panels in series --->>> (0.1766) * 8 = -1.413 V/deg. C. I'd suspect the greater value I got, besides coming from less than perfect measuring and instruments, might be due to any temp. induced increase in voltage drop from wiring and other things in the system on the roof and on the way to the inverter. All in all, those series of 60 measurements, and the rest of the measurements over the last 3+ years have given me confidence that theory as in the open literature matches reality to a pretty good degree.

FWIW to those interested (Sensij ?), and a bit off topic, I also have data for all those measurements and all of the other measuring days (which, by my design were all clear sky days, at least +/- ~ 1/2 hr. around the time of minimum incidence angle for any measuring days, which now amount to several hundred) comparing the Sandia method of estimating panel temps. with actual measured panel temps.

Bottom line on the Sandia method : I got quite good agreement between measured ave. array temps. and the results of the Sandia PVArray Performance Model using eq. 11 & 12, with a = -3.56, b= -0.075 and delta T = 3 C.)

A couple of details on that: The measured individual panel temp. distributions over the array, not surprisingly, follow reasonably well with established heat transfer correlations for temperature distributions for the condition of gas flow (air) over a flat plate as can be derived from heat transfer and boundary layer theory to a reasonable degree given the uncertainty introduced by the uncontrollable and variable effects of a constantly changing wind vector, and also the uneven underside of the array. Bottom line: For the 60 measuring days my measured temperatures agree very well with the Sandia method , eq. 11 and 12 which resulted in an average array temp. for the 60 measuring days of 52.4 C. for both the average measured array temp. and the estimated array temp. using the Sandia method. The average difference in the measured vs, calc'd temps using all 60 data taking trials was 0.05 C., with a std. dev. of 1.7 C. I attribute the rather wide or large std. dev. of the difference between measured and calc'd array temps. being an indication of the uncertainty introduced by the wind.

Max: If you are still reading all this mental spoor, array temp. is reasonably well correlated with array voltage, much less so for current. Array temp. is mostly a function of irradiance on the array as can be estimated and verified by an energy balance on the array. I've got data fir that as well. Dig into a few sources and you may come to the same conclusions. Until then, I'd suggest you trust the spec sheet data.

[J.P.M.]

too many assumptions- no way cells were near 46C at 8:00 AM in the morning and it doesn't matter how hot the coming day would be, more important how hot the previous night was. With 25C ambient 35C cell may be but that would give drop of only 10x0.004x30 = 1.2V per panel. I can easily see low irradiance to be prevailing factor as my own array produces < 5% power around 8:00 AM from its middady peak and 20% irradiance level already gives 5-7V drop. Of course roof orientation and local weather patterns play big role here.

Max: I'd agree that 46 C. at 8 A.M. seems too high, but the measurement conditions' accuracy also seems a bit sketchy. I did a couple of SWAGS using the given the conditions and got closer to ~ 35 - 40 C. as a back plate temp. for clear sky conditions for that location, date, time and ambient temp., or ~ 12 C. or so above ambient. Still, front surface temps. can be higher by maybe 5 deg. c. or so, and repeating the idea that accuracy of reporting may not be the best, including the thermometers, I'd take the claimed numbers with a grain of salt and a skeptical eye, but still possible.

As for what the ambient temps. were at night, and how they might depress panel temps: Since most panels have a thermal time constant measured in minutes, I'm pretty sure any nocturnal cooling effects would have dissipated by 0800 the next morning. Look up something called the Biot number as it relates to heat dissipation.

[max2k]

Max: I'd agree that 46 C. at 8 A.M. seems too high, but the measurement conditions' accuracy also seems a bit sketchy. I did a couple of SWAGS using the given the conditions and got closer to ~ 35 - 40 C. as a back plate temp. for clear sky conditions for that location, date, time and ambient temp., or ~ 12 C. or so above ambient. Still, front surface temps. can be higher by maybe 5 deg. c. or so, and repeating the idea that accuracy of reporting may not be the best, including the thermometers, I'd take the claimed numbers with a grain of salt and a skeptical eye, but still possible.

As for what the ambient temps. were at night, and how they might depress panel temps: Since most panels have a thermal time constant measured in minutes, I'm pretty sure any nocturnal cooling effects would have dissipated by 0800 the next morning. Look up something called the Biot number as it relates to heat dissipation.

I mentioned the 'previous night' referring to the fact that since ambient temp was 25C it was very unlikely cells / roof were heated up during the night so all we had was current irradiation at the time of measurement. At 8:00 AM I don't expect it to be around 300 W/m2 but I'm first to admit I don't know for sure. Given time constant in minutes I also don't expect front of the cell to be hotter than back by that much at 8:00 AM.. If OP posted his power production curve on that day it would be easy to see. This whole discussion actually revolves around the fact how strong the irradiation was at the time of Voc test. For all we know they could have passing clouds or 100% clear skies so production power curve would help a lot.

All I'm saying is I was surprised to learn that at least for my own location Voc in the morning will be mostly affected by low irradiance.

[max2k]

I've got something to add to the connection between panel or array temperatures as measured, and theory as it currently exists. FWIW, I've used theory, a lot of it from the very same sources Sensij quoted from to develop a methodology to correlate array temperatures and array voltages. I needed to develop that to be able to estimate array fouling and investigate how array fouling changes as f(time, weather conditions). I then tested the method by measuring each of my array's 16 panel's temps. in 4 random places on each panel over a 6 to 8 minute time period both before and after the time of minimum incidence angle (16 X 4 X2 = 128 temp. measurements per day's trial, all averaged together and used as that day's average measured temp.) Then, at the time of min. incidence angle for each of those days, and between measurements, I recorded string voltages and current at the inverter input, string power D.C. inputs to the inverter, and total A.C. output from the inverter for well over 100 clear days. I winnowed those 100 + measurement days down to 60 days of such measurements split about evenly between winter and the following summer, with all days in each season being consecutive clear days. In that sense, beyond the start and stop of each season and the goal of ~ 30 measuring days in each season, I didn't cherry pick the data. I also used on site weather data, including GHI, taken continuously in one minute intervals from a Davis instruments Pro II Plus weather station located about 4 ft. north of the array at the same elevation as the highest array point.

A couple of results: Array or sting current correlates very well with POA irradiance. After confirming the d(current)/d(irradiance) ~ = +.0035 A/deg. as per the spec sheet (but that was hard to measure and I can't confirm it beyond saying it isn't much), I got a current change per irradiance change based on calculated POA irradiance of ~ 164-169 W/A-m^2), vs. a spec sheet value of (1,000 W/m^2)/5.98 Amp = 167.2 W/(A-m^2). I'd suggest that if it can be compared to and calibrated against a pyranometer, an array's output can be used as it's own pyranometer.

Bottom line: Theory as I understand it matches most all the data I've taken, and vice versa to a better degree than I had expected. I believe I confirmed the spec sheet claims reasonably well, including not only that current does not vary too much with temperature but directly with POA irradiance, but also that the second derivative of voltage with respect to temperature is close to zero, at least for the array temps. I measured (33.5 C- 67.1 C.). I did some farting around with the numbers, just for the hell of it, and SWAGed 2d deriv. value of volt/temp. of ~ -2.4 EE-4, but that's probably no more than B.S. that's not defensible given the instruments I'm using, but I had fun doing it and refreshed some math along the way.

One semi detailed example of what I got from all these measurements: String Voltage increment per degree of array temp. difference worked out to have an average value of -1.477 v/deg. C ( std. dev. = 0.123V/deg. C, N=60). The spec sheet per deg. C Voltage drop is listed as -0.1766V/deg. C. per panel. --->>> 8 panels in series --->>> (0.1766) * 8 = -1.413 V/deg. C. I'd suspect the greater value I got, besides coming from less than perfect measuring and instruments, might be due to any temp. induced increase in voltage drop from wiring and other things in the system on the roof and on the way to the inverter. All in all, those series of 60 measurements, and the rest of the measurements over the last 3+ years have given me confidence that theory as in the open literature matches reality to a pretty good degree.

FWIW to those interested (Sensij ?), and a bit off topic, I also have data for all those measurements and all of the other measuring days (which, by my design were all clear sky days, at least +/- ~ 1/2 hr. around the time of minimum incidence angle for any measuring days, which now amount to several hundred) comparing the Sandia method of estimating panel temps. with actual measured panel temps.

Bottom line on the Sandia method : I got quite good agreement between measured ave. array temps. and the results of the Sandia PVArray Performance Model using eq. 11 & 12, with a = -3.56, b= -0.075 and delta T = 3 C.)

A couple of details on that: The measured individual panel temp. distributions over the array, not surprisingly, follow reasonably well with established heat transfer correlations for temperature distributions for the condition of gas flow (air) over a flat plate as can be derived from heat transfer and boundary layer theory to a reasonable degree given the uncertainty introduced by the uncontrollable and variable effects of a constantly changing wind vector, and also the uneven underside of the array. Bottom line: For the 60 measuring days my measured temperatures agree very well with the Sandia method , eq. 11 and 12 which resulted in an average array temp. for the 60 measuring days of 52.4 C. for both the average measured array temp. and the estimated array temp. using the Sandia method. The average difference in the measured vs, calc'd temps using all 60 data taking trials was 0.05 C., with a std. dev. of 1.7 C. I attribute the rather wide or large std. dev. of the difference between measured and calc'd array temps. being an indication of the uncertainty introduced by the wind.

Max: If you are still reading all this mental spoor, array temp. is reasonably well correlated with array voltage, much less so for current. Array temp. is mostly a function of irradiance on the array as can be estimated and verified by an energy balance on the array. I've got data fir that as well. Dig into a few sources and you may come to the same conclusions. Until then, I'd suggest you trust the spec sheet data.

thank you for so detailed research. I think you made a typo there: d(current)/d(irradiance) ~ = +.0035 A/deg. was most likely meant to be d(current)/d(temp) ~ = +.0035 A/deg.
second derivative being close to 0 means the first one was close to const meaning the original function was linear so we can rely on coefficient listed in the data sheet
I found your measurement results/accuracy quite satisfactory as my expectation is if anything fits within +- 10% it is already good enough for 90% of applications. I suspect your std dev was related to the temp itself and not difference between measured/calculated values (0.05 C., with a std. dev. of 1.7 C). If that's the case I think you have very good model within data range it covers.

I tend to take the models with grain of salt especially when actual data are available. What often happens some set of data is taken, model is built and then the model is extrapolated beyond original data and conclusions are made based on those results which may or may not relate to any reality. I think this is what we're dealing with here as we're trying to analyze OP system behavior extrapolating models built at higher irradiance levels. I was burned more than once doing this .

[sensij]

I think this is what we're dealing with here as we're trying to analyze OP system behavior extrapolating models built at higher irradiance levels. I was burned more than once doing this .

Perhaps if you would read the published material in which the models are derived and validated, you wouldn't get burned by misusing them. There is quite a bit of information out there describing when the single diode / CEC performance models are appropriate. Your apparent unwillingness to respect the extensive work that has gone into their development is making this discussion much more difficult than it needs to be.

[J.P.M.]

thank you for so detailed research. I think you made a typo there: d(current)/d(irradiance) ~ = +.0035 A/deg. was most likely meant to be d(current)/d(temp) ~ = +.0035 A/deg.
second derivative being close to 0 means the first one was close to const meaning the original function was linear so we can rely on coefficient listed in the data sheet
I found your measurement results/accuracy quite satisfactory as my expectation is if anything fits within +- 10% it is already good enough for 90% of applications. I suspect your std dev was related to the temp itself and not difference between measured/calculated values (0.05 C., with a std. dev. of 1.7 C). If that's the case I think you have very good model within data range it covers.

I tend to take the models with grain of salt especially when actual data are available. What often happens some set of data is taken, model is built and then the model is extrapolated beyond original data and conclusions are made based on those results which may or may not relate to any reality. I think this is what we're dealing with here as we're trying to analyze OP system behavior extrapolating models built at higher irradiance levels. I was burned more than once doing this .

No thanks required. I did it for my own purposes.

You are correct. I made an error, except it wasn't a typo, it was brain flatulence. Should have been d(current)/d(temp.), NOT d(current)/d(irradiance). And yes, for the most part, and without suffering a measureable error in accuracy, the coefficients of voltage and current as f(cell temp.) over common working array temps are, for all practical purposes, constant. The locus of Mpp's on a V vs. current current graph is also mostly a straight line in ranges of working interest.

On the 1.7 std. dev. #, that was indeed 1.7. That 1.7 is the std. dev. of the average of 60 measurements of (average of 128 temp. measurements - Sandia method temp. calculation). The means are close but the standard dev. is relatively wide as is the range = 9.0 C. A bit more info: I believe the thing messing up the range of differences in measured vs. calc'd temps. is how the Sandia method treats the wind vector contribution to the array temp. I believe it overstates it a bit. To get an average wind velocity, I use the average of the wind velocity and wind gust for each of the 7 minutes around the measurement minute and average those 7 numbers. I use 7 minutes because that's about the time length of 2 thermal time constants for the array as I estimate it under an average wind velocity of 1.8 m/sec. which is the average wind velocity since I put the weather station on the roof. It's also close to the amount of time it takes for 1/2 of the temp. measurements. I designed my array so I could get under it to measure the backside of the panels, but for the 12 - 16 minutes or so around the inverter measurements, which take 1 minute + 90 sec. travel time and 30 sec. to scratch my junk, I'm either under the array or running to the garage to record inverter display readings. So, for those times, there's a lot of asses & elbows. Anyway, on the Sandia wind situation, what I think is going on is that the Sandia method tends to calc wind as having more of an affect on plate temp. than I'm measuring. Looking at the data, days with higher wind velocities tend to have lower Sandia method values for array temp. than measured values, with days having lower wind velocities tending to give higher plate temps. than I'm measuring. I considered backing into a Sandia temp. coeff. that made the Sandia data a closer fit to measured, but have not done so. After 300+ measurement days, I'm not changing it. Besides, the wind is the most uncertain variable in these types of analyses anyway. for all I know, my measuring method might suck and the Sandia method be spot on. My measurement method isn't the best, but others I've seen are no better and a lot are worse.

[J.P.M.]

I mentioned the 'previous night' referring to the fact that since ambient temp was 25C it was very unlikely cells / roof were heated up during the night so all we had was current irradiation at the time of measurement. At 8:00 AM I don't expect it to be around 300 W/m2 but I'm first to admit I don't know for sure. Given time constant in minutes I also don't expect front of the cell to be hotter than back by that much at 8:00 AM.. If OP posted his power production curve on that day it would be easy to see. This whole discussion actually revolves around the fact how strong the irradiation was at the time of Voc test. For all we know they could have passing clouds or 100% clear skies so production power curve would help a lot.

All I'm saying is I was surprised to learn that at least for my own location Voc in the morning will be mostly affected by low irradiance.

I plugged data for the location of the array into the Bird clear sky model for GHI (at 0700 std. time, then used my own modeling to get a POA values and got about 360 W/m^2. PVWatts for a clear day seems to give ~ 370 or thereabouts so 365 W/m^2 seems possible, or at lest defensible. As things go, I'd think the front side glazing would be relatively warm as would the cells, but I'm not about to do an energy balance on the array nor a transient temp. profile to find out. But, all the irradiance info and the idea that a heat loss coeff. of very roughly 25 - 35, say, 30 W/M^2 deg. C. where the area is that of one side of the array or panel with the heat loss taken as from both sides might be a reasonable # by my reckoning, reading and experience. If so, the array temp. might be something like T, amb. + 360/30 = 25 + 12 = 37 C., which is what I wrote earlier. The idea that the data may not be the best, or the temps. and times different, all add to the confusion. Kind of a GIGO situation.

As for your Voc, it is affected by low panel temps. The low panel temps are the result of low irradiance. As discussed in a prior post, Voltage is mostly and all practical purposes a function of cell temp. The higher the cell temp., the lower the Voltage. An energy balance is required to find the array temp. Such an energy balance will result in a lower cell temp. at times of low irradiance as in the morning, leading to higher Voltage.

Experiment: If you use a solar simulator to irradiate a panel at 500 W/M^2 while holding the cell temp constant at, say 25 C, via some reliable and sophisticated cooling method you would measure a certain voltage. If you then heat the cell to a temp. of, say, 65 C, and hold that temp. constant while also keeping the irradiance constant at that same 500 W/m^2, and again measured Voltage the same way as the first trial, you would measure a lower voltage. The irradiance has not changed, nor have any of the other conditions. Only the cell temp. has changed, but the voltage has dropped. Conclusion: Irradiance does not affect Voltage. Temperature does affect Voltage. Been done many times. I've done It - long ago and far away.

[max2k]

Perhaps if you would read the published material in which the models are derived and validated, you wouldn't get burned by misusing them. There is quite a bit of information out there describing when the single diode / CEC performance models are appropriate. Your apparent unwillingness to respect the extensive work that has gone into their development is making this discussion much more difficult than it needs to be.

why bother with models when direct IV curve from the data sheet is available? The only problem is we know neither irradiance level (no power output curve is available for that day) nor exact temperature. At this point we can all respectfully disagree on the conclusion and wait until someone with more reliable data comes along.

[max2k]

As for your Voc, it is affected by low panel temps. The low panel temps are the result of low irradiance. As discussed in a prior post, Voltage is mostly and all practical purposes a function of cell temp. The higher the cell temp., the lower the Voltage. An energy balance is required to find the array temp. Such an energy balance will result in a lower cell temp. at times of low irradiance as in the morning, leading to higher Voltage.

right and OP problem was low Voc so either temp was unusually high (for me) or irradiance low.

Experiment: If you use a solar simulator to irradiate a panel at 500 W/M^2 while holding the cell temp constant at, say 25 C, via some reliable and sophisticated cooling method you would measure a certain voltage. If you then heat the cell to a temp. of, say, 65 C, and hold that temp. constant while also keeping the irradiance constant at that same 500 W/m^2, and again measured Voltage the same way as the first trial, you would measure a lower voltage. The irradiance has not changed, nor have any of the other conditions. Only the cell temp. has changed, but the voltage has dropped. Conclusion: Irradiance does not affect Voltage. Temperature does affect Voltage. Been done many times. I've done It - long ago and far away.

how would you explain right side of IV curves where they intersect I = 0 level? They clearly show irradiance influence on Voc.

BTW, going completely scientific about this when measuring cell temperature what should be considered cell temperature? I think back of the cell is approximation down while front- approximation up as the actual element is behind layer of glass at least. I understand that energy comes from the front and then dissipates in all directions and some of it gets converted cooling cell a bit I just wonder how big those effects are- fraction of degrees or few whole degrees? Did you come across data where someone measured those temps using infrared thermometers? Those should provide some insight as it would be relatively hard to measure temp of the front of the cell with some sort of contact sensor without influencing it if we talk about fraction of degree here - the temp probe would be in the path of the incoming energy flow shading the cell it is intended to measure.

[sensij]

why bother with models when direct IV curve from the data sheet is available? .

Believe it or not, one of the outputs of the CEC testing is to determine the correction factors needed to reconcile the datasheet values with actual 3rd party test results under more than just STC or PTC conditions. Those correction factors allow the panel's performance to be estimated more reliably in the real world than estimates produced from the datasheet information alone, and are most useful when employed within the model for which they were validated.

[max2k]

Believe it or not, one of the outputs of the CEC testing is to determine the correction factors needed to reconcile the datasheet values with actual 3rd party test results under more than just STC or PTC conditions. Those correction factors allow the panel's performance to be estimated more reliably in the real world than estimates produced from the datasheet information alone, and are most useful when employed within the model for which they were validated.

ok, my last argument, I promise - how would you explain the area of the IV curves from the data sheets where they intersect V axis on the right as not a dependency of Voc on the irradiance? It is not negligible, in 5-7V range. IMO we should treat them at face value and conclude the Voc depends on the irradiance no matter what models might say unless of course someone screwed up taking those curves and those lines don't correspond to reality. All my objections are based on those curves only, I haven't measured single thing in PV myself.

[sensij]

ok, my last argument, I promise - how would you explain the area of the IV curves from the data sheets where they intersect V axis on the right as not a dependency of Voc on the irradiance? It is not negligible, in 5-7V range. IMO we should treat them at face value and conclude the Voc depends on the irradiance no matter what models might say unless of course someone screwed up taking those curves and those lines don't correspond to reality. All my objections are based on those curves only, I haven't measured single thing in PV myself.

Look at the datasheet for the OP's panels that I linked in this post.

According to the IV curves you want to rely on, the Voc change from 1000 W/m2 to 200 W/m2 at 25 deg C is about 2.4 V, the change from 1000 to 400 is only 1.4 V. The Voc change at 1000 W/m2 from 25 deg C to 50 deg C at 1000 W/m2 is 3.4 V. I have no idea where you are coming up with 5-7 V in this context.

A reasonable guess at the conditions at which the OP's Voc measurement was taken (as well explained by J.P.M.) is something like 37 deg cell temp and 365 W/m2 of irradiance. No, we don't have the actual power curve, but looking at PVoutput data for systems in the same zip code as the OP shows clear days continuously from 6/26/17 through 7/23/17, and the timing of this thread suggests that the measurement was made sometime in that window.

Those reasonable guess conditions are not directly represented by any of what is shown on the data sheet. A model is needed to translate the lab generated data sheet information to the real world conditions in which the panel is used. That is what PVWatts, SAM, Aurora, etc do, and why those models (or something like them) is the basis for the sizing activity that is done during the planning stages.

The detailed model I posted earlier agrees with a small voltage change in response to irradiance. It agrees with a stronger voltage change in response to temperature. At the level of detail necessary for most forum discussions, it is safe to just say that in the real world voltage = f(temp) and leave it at that. At the level of detail we've gone into here, we can try to identify the likely contribution of both temperature and irradiance, but it will take more modeling time than I've put into it so far.