Comparing the Equinoxes from a Solar Production Standpoint

[NYHeel]

This is probably a dumb question but I've been wondering why the fall season is so much worse than the spring season from a solar production standpoint. I always thought that the sun pattern was basically the same on both the autumnal and vernal equinoxes and therefore I'd think that the solar production would be similar. Yet, based on both PVWatts estimates and 4+years of actual experience I can say that, at least in the NYC area, the vernal equinox is far more productive than the autumnal equinox. The expectation varies by about 10% and in practice, the good days in late March will beat the good days in late September by over 20%. Anyone know why that is if the sun's pattern in the horizon is the same and the distance from the sun is the same?

I realize that some of that difference could be due to specific issues that pertain to my array like more leaves for shading in September than March. But that should be fairly minimal and I still see significant differences in the peak production during midday at these two times.

[FFE]

Cooler cells produce more watts.

[J.P.M.]

While the solar geometry and angular inputs are mostly (but not entirely) symmetric with respect to the solstices, the weather, including clearness indices (a measure of "how cloudy" it is), and temps. are not. What you describe with respect to performance is not at all uncommon, particularly in the NE U.S. where, after trying to account for the greater (short term) variability of the weather by using the greater "long term" reliability (on a statistical basis) of the climate, spring tends to be cooler than fall for corresponding dates equidistant from the summer solstice. The historical clearness indices for spring and fall in Central Park are approx. equal, but the air MAY be a bit "clearer" in the spring with less moisture in the air due to wind patterns and lower dew points.

Leaves, debris, array cleaning schedules, shading and a lot of array specific variables will have effects on annual performance that'll difficult for models to handle, making most models inadequate to that task beyond educated SWAGs..

BTW, and a dirty little secret revealed to those willing to dig a bit: While for the most part, PVWatts, and other models that use the National Solar Radiation Data Base (the "NSRDB") and its variants seems to do a good job with respect to agreement between modeled and actual array performance, most of that weather data itself is modeled and estimated/extrapolated from limited surface and some satellite measurements. After monitoring my system for almost 8 years, the modeled performance of PVWatts and its big brother SAM is within the published tolerances both models.

Welcome to the neighborhood.

[bcroe]

I've been wondering why the fall season is so much worse than the spring season from a solar production standpoint.

In these parts (61084) the problem in Dec is dominating clouds, while about Feb some
pretty sunny days start happening. Bruce Roe

[chrisski]

This is from my Renogy:
'=================='

Operating Module Temperature -40°F to 176°F
Nominal Operating Cell Temerature (NOCT) 47±2ºC
Temperature Coefficient of Pmax 0.23%/ºC
Temperature Coefficient of Voc 0.33%/ºC
Temperature Coefficient of Isc 0.05%/ºC

So, it has got to be more than just temperature coefficients. My first guess would be weather patterns.

I also noticed this year that on my September equinox walk, the sun did not rise above the mountains at 090 like I expected. It was a little to the north of that. I saw that day was 20 minutes longer than night. That lead to me looking it up on Google and seeing that equinox is really time that the earth passes at 270 and 090 through it's rotation around the sun. The data for my area: https://www.wunderground.com/history...date/2020-9-22

When I was a middle school teacher, we used to teach that equinox day and night were the same, but I guess that's not quite right. I got as far as deciding it's the 90° and 270° in the path around the sun and must have something to do with the elliptical orbit. Also seems the earth is slightly closer to the sun in the fall than in the winter. https://socratic.org/questions/how-f...52.1%20M%20km.

My gut feeling is the efficiency difference has far more to do with weather patterns than a 20 minute difference to the day a four or five degrees centigrade. Although the math is beyond me, I doubt distance from the sun and slight temperature variation adds up to 20%.

[peakbagger]

I agree on the temperature aspect making big difference with snow on the ground another factor. March 21st is generally a lot cooler than September 21st. I usually still have snow on the ground in front of the panels in march and get some boost from light reflected off the snow. I manually adjust my panel angle on my pole mount and usually leave the panels at winter angle (30 degrees) for a few weeks after march 21st until the threat of major snowfalls reduce yet even with the steeper less than optimal panel angle I still produce more as the sun reflecting off the snow compensates for the steeper angle. September 21st may be 50 degrees F warmer with no snow cover so I get hit with warmer temps derating the panel and lower incoming light. The tradeoff is in my location the spring weather pattern runs roughly on 3.5 day cycle so I might get one sunny day two partially sunny days and one snowy day. In September I am usually in a stretch of multiday high pressure systems and may go without rain for several weeks. The shift to fall pattern usually occurs a week or two after Sept 21st.

Its all academic to me, With the exception of changing the array angles seasonally, I pretty well ignore production these days, as long as the green LEDs are lit up on the inverters I am happy. I do have to manually report my meter reading monthly for SRECS so I do see a monthly total generation and if its significantly off then I might go looking for cause.

[NYHeel]

While the solar geometry and angular inputs are mostly (but not entirely) symmetric with respect to the solstices, the weather, including clearness indices (a measure of "how cloudy" it is), and temps. are not. What you describe with respect to performance is not at all uncommon, particularly in the NE U.S. where, after trying to account for the greater (short term) variability of the weather by using the greater "long term" reliability (on a statistical basis) of the climate, spring tends to be cooler than fall for corresponding dates equidistant from the summer solstice. The historical clearness indices for spring and fall in Central Park are approx. equal, but the air MAY be a bit "clearer" in the spring with less moisture in the air due to wind patterns and lower dew points.

Leaves, debris, array cleaning schedules, shading and a lot of array specific variables will have effects on annual performance that'll difficult for models to handle, making most models inadequate to that task beyond educated SWAGs..

BTW, and a dirty little secret revealed to those willing to dig a bit: While for the most part, PVWatts, and other models that use the National Solar Radiation Data Base (the "NSRDB") and its variants seems to do a good job with respect to agreement between modeled and actual array performance, most of that weather data itself is modeled and estimated/extrapolated from limited surface and some satellite measurements. After monitoring my system for almost 8 years, the modeled performance of PVWatts and its big brother SAM is within the published tolerances both models.

Welcome to the neighborhood.

Thanks for the insight. In my particular case the PVWatts estimates haven't been very accurate. I've produced about 10% lower for the 4+ years I've had solar. It's usually the fall and winter months where I really under produce though spring and summer aren't great either. This is using what I believe are accurate measurements of tilt and azimuth. I used a conservative 14% for system losses. Of course I realize snow plays a role but even in years with very little snow (winter 2019/20 had almost no snow in the NYC area) my falls and winters (more so winters) are very poor. But I realize weather works in patterns and cycles so it's always possible that the last 4 years have been poor for solar production.

But I've also found the patterns of PVWatts to be a little off. For example, they estimate April to be a bit better (by 3%) than May but through 4 years May has outproduced April by 10% on a per day basis. Also, August has been my best month but PVWatts has it as my 5th best month.Of course I realize that 4+ years is still highly anecdotal but it does make me wonder a bit if their weather data is still fairly accurate.

The good news is that this past spring was very good with very little rain during May and June. Those two months were the best months I've had compared to expectations since the first month I went active. They were also the first time I've had consecutive months above PVWatts expectations since the first two months. Actually only 1 of the previous 31 months were above PVWatts expectations at that point. But that's not really the point of my original post.

[NYHeel]

This is from my Renogy:
'=================='

Operating Module Temperature -40°F to 176°F
Nominal Operating Cell Temerature (NOCT) 47±2ºC
Temperature Coefficient of Pmax 0.23%/ºC
Temperature Coefficient of Voc 0.33%/ºC
Temperature Coefficient of Isc 0.05%/ºC

So, it has got to be more than just temperature coefficients. My first guess would be weather patterns.

My gut feeling is the efficiency difference has far more to do with weather patterns than a 20 minute difference to the day a four or five degrees centigrade. Although the math is beyond me, I doubt distance from the sun and slight temperature variation adds up to 20%.

I guess weather is the only explanation if the sun patterns are basically the same. I've tracked every day since I went active in a spreadsheet using a revenue grade meter to track my production. For a 4 year sample size of the period from 3/20-3/31, I've produced as much as 67 kWh (twice), at least 60 kWh 7 times and at least 50 kWh 17 times. Yet during the same 4 year period using 9/20-10/1 I've produced a maximum of 55 kWh and at least 50 kWh only 5 times. You'd think on a clear day with the two periods would at least be close to each other when adjusting for temperature differences. Even looking at the PVWatts estimates, they have March as a slightly better month than September which you would not expect as most of March is still winter and most of September is still fall.

[J.P.M.]

I guess weather is the only explanation if the sun patterns are basically the same. I've tracked every day since I went active in a spreadsheet using a revenue grade meter to track my production. For a 4 year sample size of the period from 3/20-3/31, I've produced as much as 67 kWh (twice), at least 60 kWh 7 times and at least 50 kWh 17 times. Yet during the same 4 year period using 9/20-10/1 I've produced a maximum of 55 kWh and at least 50 kWh only 5 times. You'd think on a clear day with the two periods would at least be close to each other when adjusting for temperature differences. Even looking at the PVWatts estimates, they have March as a slightly better month than September which you would not expect as most of March is still winter and most of September is still fall.

If you track the Global Horizontal Insolation ("GHI") at the array's site (using a pyranometer), convert that to Plane of Array Insolation ("POA") and then compare that POA insolation to your array's daily production, you'll find that's a pretty strong correlation that's reliable enough to get a decent first appox. of the effects that other climate variables have on cell temps. and so array output.

As for weather vs. climate, when I got zapped by the alternate energy bug in the mid '70's, it took me a while to appreciate the wisdom of the old saw: Climate's what you expect - weather is what you get.

Climate is more concerned with long term averages - it's warmer in summer than winter, that sort of thing. You can get a mean and standard deviation for any climate variable for any day or string of days over 30 years. That's an example of climate. However, that info is mostly useless in predicting what a single unique day's ave. temp. will be 6 months from now. That's weather.

Any comparison of an array's output for any one day to what a model such as PVWatts may produce for that day will have a very low probability of being close to the long term average. That's a consequence of the chaotic nature of weather.
Any consecutive 30 days average of a weather variable for any one year will most likely be a lot closer to the long term 30 day variable for those same dates. That's the nature of climate. It's less chaotic.

My daily data for what PVWatts models for my array divided by what my array actually produced for the same dates daily over close to 7 years :
Average of (Actual single day output / PVWatts modeled single day output) = 1.169, std. deviation = 0.893. Minimum = 0.027. Maximum = 12.1. n = 2,544.

The corresponding 31 day totals for the same period :
Average of (31 day actual total array output / PVWatts modeled 31 day total output) = 0.997, std. dev. = 0.100. Minimum = 0.646. Maximum = 1.30. n = 2,513.

For running 365 day periods:
Average of (365 day running actual total array output / PVWatts modeled 365 day running total array output) = 0.996, std. dev. = 0.021. Minimum = 0.9575. Maximum = 1.053. n = 2,179.

As the PVWatts info screens state, (I'd assume) once the model is dialed in, any comparison of an unshaded and operating array's total 31 day output to what PVWatts models suggest for an array's output for the same period will probably be ~ +/-30 % or so. Also as stated in the model's help/info screens, annual modeled output will be up to +/- 10 %. As described above, my experience seems to agree with those estimates.

See the PVWatts info screens for a better discussion.