Rain - Best For Cleaning Solar Panels?

[Ian S]

I was just getting ready to clean my panels a week or so ago but with rain forecast, I decided not to. In the past,several of us here have attempted some SWAGs to get an idea of how much cleaning the panels gets you in terms of improved output. IIRC, the general consensus was somewhere between 4% and 6%. A lot depends on just how dirty the panels are of course. This past winter, we've had some rain here in Phoenix and so I haven't felt much need to get up on the roof and clean the panels myself. Anyway, a system moved through last weekend starting with clouds around the 5th then with just under a quarter inch of rain in the wee hours of the 8th. Then on the 10th we had some short but heavy showers move through and drop another quarter of an inch of rain. Prior to the system moving through, we had about five straight days of clear blue skies and sunshine. Starting on the 11th we had four days pretty much similar with blue skies and sunshine. Average mean temperatures for the days following the rain were about 5 deg. F higher than for the days prior to the rain. The results are shown below but the output after the rain is 13.4% higher than before the rain. That's a pretty significant jump. It doesn't appear to be due to temperature difference and I don't recall the winds being significantly different during the before days vs the after days. Maybe cleaner air after the rain? Maybe my panels were extra dirty this time. I really don't know.

[J.P.M.]

Ian: Nice to see you.

Some work I've done has led me to the preliminary conclusions that get reinforced the more I measure, record and fiddle with my array fouling investigations.

I've no trouble believing your observations. Briefly, my array seems to fouls at a rate of something like 1 % per week when it doesn't rain. LOTS of modifiers to that. If there is a Santa Ana wind event, the array will foul sooner and a lot more. If it rains, again, very roughly, and depending on length and intensity of the rain, something like 2/3 or more of the extant fouling seems to be removed. The fouling rate then continues at about the 1 % per week rate and is additive to what was not removed by the rain.

I continue to be suspicious that the RATE of panel/array fouling is asymptotic in nature. However, I have not yet been able to detect such behavior as the longest dry stretch I've had since starting this odyssey ( ~ May, 20104) is about 7 weeks. However, common sense and experience with skylights leads me to believe that 6 months without rain, for example, will not reduce transmittance by 26 %.

Unfortunately or not, the nature of the phenomena associated with fouling is not any more predictable than the weather. Trying to detect a small difference in output - say 1/2% fouling, even when measuring the major variables like irradiance, wind vector and plate (panel) temps., and a bunch of other ones as well, is mostly an exercise in futility without many repeated sets of data. Even then, about the best I can do is guess at a trend, based on many measuring events that are as close to identical in procedure as I can make them, and then look for trends in the data. I'd estimate the best I can do on guessing panel fouling from any one measuring event is probably +/- 1.5 % reduction in output. That is, if the actual fouling is 2%, it might be something like 0.5 to 3.5%. Close to 200 or so measuring events may allow me to give what amounts to an educated guess that, on average, my array fouls at about a 1% per week rate, subject to a lot of variation, and rain removes about 1/2 to 3/4 of the accumulated gunk.

You're being in the desert and in Phoenix which might probably mean your rate of fouling will be different, I'd guess more than mine, but I could be wrong. I've confirmed to myself at least, is that this is a crazy business, highly variable and difficult to quantify.

[foo1bar]

Maybe cleaner air after the rain? Maybe my panels were extra dirty this time. I really don't know.

So 4/1 to 4/14 is ~13% better.
What's your delta form 4/14 to 5/1?
Because you also have more daylight because you're 2 weeks closer to summer.
And that can definitely also affect your output.
Looking at 4/14 -> 5/1 (or around then) should help you see how much difference is due to more sunlight.
I'm sure it's only a minor impact - but it probably accounts for some of that ~13%

[azdave]

Since I'm also in the Phoenix area, out of curiosity, I compared my numbers to yours. I show that after the rain I saw about a 3% increase. I certainly understand that this is a very unscientific way of looking at the difference between dirty and clean panels but my results still assure me there is no reason to be up there washing them. Before the rain on April 8th we had not had any rain since late January and from the ground the panels looked pretty filthy to me.

After throwing in some wind and temperature errors and length of day adjustments I still can't see it ever being more than 5% gain after cleaning. This time of year I'm only going to get wholesale rate on what I bank (I'm with SRP) so no washing for me unless it's some rare event like the dust storm of July 2011 that left 1/4 of red powder coating everything in the area.

The graph below is from SRP and shows the average daily temp over the generation graph. The actual daily collection data I used came from my inverter logging which is more detailed.

[J.P.M.]

This array fouling business is probably one of the most murky areas of alternate energy which is almost by definition an inexact endeavor in the best of circumstances.

I've tried to remove as much uncertainty as possible under my circumstances by taking data as a " snapshot" at the time of a day's min. incidence angle on my array, and compare that measured power output to what I estimate the power the array might produce if it was absolutely clean under the same conditions of measured GHI, calc'd POA, recorded temp, wind vector and average array temp. and a bunch of other stuff like dew point, roof deck temp., both exposed to the sky and under the array as well as all the stuff that the inverter has like voltages, currents and power in/out o the inverter, etc.

It's hoped that doing that enough times, and with precision in the method and the instrument's precision, will make up for some of the uncertainty in the instrument's and the experimenter's accuracy. Hope springs eternal.

Over time, and with enough care, I believe I've noticed trends that seem to make some sense, but there are always surprises that make me say "WTF ?".

My 1%/week fouling rate per week - if it doesn't rain - and the conclusion based on looking at the data that rain removes maybe ~~ 1/2 to maybe 2/3 or so of the accumulated crud depending on rain intensity and duration, seems to make sense given the data and results I've obtained. But any week or day for that matter can be way off of that #, depending on the weather.

I'd add that working # of ~ 1% fouling/Week without rain, to the extent it's representative of some version of reality at all, is probably only good for my array. I think it may be representative of fouling rates for other arrays in my neighborhood from some cursory comparison with other arrays neighbors allow me to cop a scope at, and maybe loosely applicable to north county San Diego in a qualitative sense.

For those perhaps interested, I'm still trying to make sense of whether using array voltages can be an alternate method of estimating array temps., different from actually getting under the array and taking what amounts to 128 separate measurements with an IR thermometer per measuring event, which is, in my case possible, but a real PITA. And if possible, what the differences in the two methods might be. Long story, but it seems the two methods give similar results +/- some. More detail soon. I take data directly from the inverter as the monitoring method that Sunpower has seems to be about 5-20 min. off in time stamp, and anyway, it's 5 min. ave, data. not real time.

[Ian S]

So 4/1 to 4/14 is ~13% better.
What's your delta form 4/14 to 5/1?
Because you also have more daylight because you're 2 weeks closer to summer.
And that can definitely also affect your output.
Looking at 4/14 -> 5/1 (or around then) should help you see how much difference is due to more sunlight.
I'm sure it's only a minor impact - but it probably accounts for some of that ~13%

Actually because I took averages over several days before and after the rain, it's probably more like a 10 day delta and sure, there will be an effect of being closer to maximum vertical incidence of the sun's rays on the panels and the longer days. I guess I was expecting somewhere between 4% and 6% change since that would be consistent with previous effects I've noticed with washing the panels. But not 13%. One thing I'd note that may have an impact vs others' results is my fairly low panel angle of 15 degrees (flat roof installation.) Might result in the dirt tending to build up faster and be less easily removed.

[Ian S]

For those perhaps interested, I'm still trying to make sense of whether using array voltages can be an alternate method of estimating array temps., different from actually getting under the array and taking what amounts to 128 separate measurements with an IR thermometer per measuring event, which is, in my case possible, but a real PITA. And if possible, what the differences in the two methods might be. Long story, but it seems the two methods give similar results +/- some. More detail soon. I take data directly from the inverter as the monitoring method that Sunpower has seems to be about 5-20 min. off in time stamp, and anyway, it's 5 min. ave, data. not real time.

Yes, J.P.M. that would be interesting if you could get reliable temperatures that way. Lately, I'm less inclined to go up on my roof unless really necessary. My next door neighbor fell from hers while cleaning the scuppers recently and was seriously injured. A flat roof, being easy to walk on, gives a false sense of security and one section of my panels gets pretty close to the edge parapet so I have to be extra careful there.

[azdave]

Yup. Just the deductible and co-pays from a twisted ankle can negate years of gains maintaining clean panels. Be careful.

[MGE]

An easy way to compare apples to apples is to wait for the panels to get dirty again then right before the next rain (In SD here you might be waiting a while) is cover up a few panels that produce comparably to each other and then after the rain uncover the few you covered up and then compare. Just an idea.
I recently just tilted 17 of my panels (south facing) from 2 degree to 25 degree and did them one row at a time to realistically compare the outputs from the old to the new angles. I figured that I would get about a 10% increase in production and thats how it worked out to. But I had the before and after producing at the same time, day, temp, clouds etc. to verify. Oh yeah I clean mine too probably once a month. Easy access from 2nd floor balcony makes it a breeze. Good luck.

[J.P.M.]

An easy way to compare apples to apples is to wait for the panels to get dirty again then right before the next rain (In SD here you might be waiting a while) is cover up a few panels that produce comparably to each other and then after the rain uncover the few you covered up and then compare. Just an idea.
I recently just tilted 17 of my panels (south facing) from 2 degree to 25 degree and did them one row at a time to realistically compare the outputs from the old to the new angles. I figured that I would get about a 10% increase in production and thats how it worked out to. But I had the before and after producing at the same time, day, temp, clouds etc. to verify. Oh yeah I clean mine too probably once a month. Easy access from 2nd floor balcony makes it a breeze. Good luck.

I figured that method or some close version of it might work with a system equipped with micros, but I haven't been able to talk any micro equipped neighbors into it yet.

I effectively do the same thing but with the entire array when I clean it and measure before/after, or the day after a rain event if that day after is clear. I try to clean the array on or around the solstices, and at various other times such as around the anniversary of the array startup for maybe some SWAG info on annual panel deterioration. I've cleaned the array 7 times, with the 4 day sequence referenced below counting as 1 very thorough cleaning.

For that cleaning, and as an example of the method, during the period 06/14/2015 to 06/21/2015, I measured array and weather parameters, array temps, voltages at the inverter and used the Sandia method to estimate array temps., all in the same and usual manner as all the measurements, all done at the time of min. incidence angle on the array for that day, 8 days in a row. I wanted to hit the equinox but the weather forecast wasn't promising so I chose not to gamble and took what I could get.

There was some rain prior on 06/11 & 06/12 - .03" & 04" respectively, and before that significant rain 05/14 to 05/27 - 1.39" total. The first set of 4 days (06/14-06/17) were measured dirty. For the second set of 4 days of measurements (06/18-06/21), I cleaned the array VERY thoroughly each of the 4 mornings with soap, a soft brush, a rinse and squeegee.

Long story short: Average estimated fouling for the first 4 days: 0.0114.That is, measured performance was 1.14% below the theoretical performance of a clean array using the same input parameters. Average estimated fouling for the 2d four days using the same logic and cleaning each of those 4 days: 0.0005, and close enough to clean in my book.

In spite of my best efforts, this is not an exact science.

I also try to clean the array a day or so after a rain, and also measure the fouling immediately after the rain event but before cleaning if that rain event is followed by a clear day. That not only allows a measure of the effect of the rain, but also any additional effect by the cleaning. I only measure on clear days, try to measure on as many clear days as possible and still have a life, with "clear" in this context meaning not only no clouds, but also stable pyranometer output with not more than a Watt/m^2 drift or maybe two in measured irradiance for a few min. around min. incidence angle/reading time.

BTW, in case anyone is wondering or cares, I clean the pyranometer sensor very carefully immediately before a reading event.

[Willaby]

I was doing some patio work last week and had the opportunity to clean just one panel, closest to the edge of the roof. First cleaning since the install in July last year. Just using car soap with a soft brush, then rinsed. That panel did appear cleaner than the rest, but none were bad. The result? Maybe 1% improved output at the most. With no shading this time of year that row of panels are within <2% of each other and the cleaned panel ranked in the middle of the group. This same result was observed each of last 3 days via the Solar Edge portal. Rains have been fairly consistent this year, with at least a mild cleaning worth each month. I doubt I will bother doing much cleaning, unless there is no rain for several months and/or they visibly need it. A good rinsing might be all that's needed.

[silversaver]

I just washing the panels from soft water hose, done. I think the water bill is higher than actual production gain.....

[SunEagle]

I just washing the panels from soft water hose, done. I think the water bill is higher than actual production gain.....

It is like trading one resource for another. Depends on which is more important or costly. With a drought water is certainly expensive to use.