Rapid, large variations in panel output on cloudy days

AceNZ
AceNZ Solar Expert Posts: 104 ✭✭✭✭
My setup includes three strings of ten 200W panels, fed into a single GT inverter. 6 kWp input, about 5.1 kWp output. All panels are facing the same direction, on a single MPPT.

I understand that when part of a panel is in shade, the output of the entire string shuts down.

What I don't understand is how panels should behave as clouds pass over. I guess I was expecting smooth ups-and-downs, like I see in the morning or evening on a clear day. Instead, what I see is rapid, large jumps from full output at 5 kW to, say, 1 kW, and then back again -- I can't tell for sure, but it looks like one or two strings may be completely turning off.

Here are links to two graphs that show what I'm talking about. First, a clear day (the dark green line is panel output):

http://pvoutput.org/intraday.jsp?id=24188&sid=22036&dt=20131223&gs=0&m=0

then today, with light high clouds:

http://pvoutput.org/intraday.jsp?id=24188&sid=22036&dt=20131226&gs=0&m=1

A few questions:

1. Is this normal? If so, why?
2. Is there anything I can do to mitigate the jumps?
3. Are micro-inverters likely to perform better on cloudy days?

Comments

  • Ralph Day
    Ralph Day Solar Expert Posts: 1,019 ✭✭✭✭
    Re: Rapid, large variations in panel output on cloudy days

    Some call it "lens-ing", but the most likely cause is the clouding reduces the photoelectric effect, production drops and the panels cool down. Cloud passes, sun comes out with a sudden increase in production on cooler panels and you get a quick spike in output. Simple physics. The colder your panels the more productive they are. My 10kw system will rarely reach 10kw production in the summer when the ambient temp is nearing 30C, but clear and sunny -30C in winter, they peak out and production is clipped at 10kw...just not enough sunny hours in the day when it's really cold :cry:.

    Ralph
  • niel
    niel Solar Expert Posts: 10,300 ✭✭✭✭
    Re: Rapid, large variations in panel output on cloudy days

    "I understand that when part of a panel is in shade, the output of the entire string shuts down."

    this is false as the bypass diodes will allow the current of the string to pass. the voltage will be a tad lower is all.
  • AceNZ
    AceNZ Solar Expert Posts: 104 ✭✭✭✭
    Re: Rapid, large variations in panel output on cloudy days
    Ralph Day wrote: »
    Some call it "lens-ing", but the most likely cause is the clouding reduces the photoelectric effect, production drops and the panels cool down. Cloud passes, sun comes out with a sudden increase in production on cooler panels and you get a quick spike in output. Simple physics. The colder your panels the more productive they are.

    I understand about the temperature effect, but I'm not sure how this explains the sudden drop in output when the light level isn't changing all that much.
    niel wrote: »
    "I understand that when part of a panel is in shade, the output of the entire string shuts down."

    this is false as the bypass diodes will allow the current of the string to pass. the voltage will be a tad lower is all.

    What I've observed is that when at least one panel in a string is more than about 20% in shadow, total power output for all three strings, as measured at the GT inverter, drops by a third -- it looks like the partly shadowed string shuts down. When the shadow moves off of the panel, full power returns.

    Are you saying that behavior isn't normal?
  • BB.
    BB. Super Moderators, Administrators Posts: 33,431 admin
    Re: Rapid, large variations in panel output on cloudy days

    "It depends".... For example, you have a "12 volt panel" (Vmp~18 volts) which has 36 series cells. And connect it to a 12 volt battery bank.

    The panel has to output >14.x volts to charge the battery bank. And each cell is Vmp~0.5 volts.

    Say the panel has two 18 cell (9 volt) strings with bypass diodes.

    You put one cell in "black". That cell is now high resistance. And no current can flow through that 9 volt sub string. Of course, there is a bypass diode around the shaded 9 volt section--And a bypass diode will have ~0.5 to 1.0 volt drop.

    The 9 volt string with ~0.5 volt diode drop will output now ~8.5 volts Vmp (at full current for that amount of sun). 8.5 volts will not charge a 12 volt battery.

    So, you don't see a "proportional" shading (1/36th of panel shading), reducing power by 1/36... If the cell is in complete blackness, you have 100% loss of power.

    Now, lets take a MPPT charge controller with 4 panels in series and a 12 volt battery bank. The MPPT controller "efficiently" takes high voltage/low current and down converts it to low voltage/high current to charge the battery bank.

    Instead of Vmp-array being 4x18 volts = 72 volts... Vmp-array with one blocked cell is now 3x18v+8.5v=62.5 volts Vmp:

    62.5 volts / 72 volts = 87% of rated power

    And, of course, if the shade covers between 1/36 to 18/36 of the same "bypass" diode string--It will not matter--The amount of power lost due to 100% shading will be the same.

    -Bill
    Near San Francisco California: 3.5kWatt Grid Tied Solar power system+small backup genset
  • AceNZ
    AceNZ Solar Expert Posts: 104 ✭✭✭✭
    Re: Rapid, large variations in panel output on cloudy days
    BB. wrote: »
    Now, lets take a MPPT charge controller with 4 panels in series and a 12 volt battery bank. The MPPT controller "efficiently" takes high voltage/low current and down converts it to low voltage/high current to charge the battery bank.

    OK so far.
    BB. wrote: »
    Instead of Vmp-array being 4x18 volts = 72 volts... Vmp-array with one blocked cell is now 3x18v+8.5v=62.5 volts Vmp:

    62.5 volts / 72 volts = 87% of rated power

    And, of course, if the shade covers between 1/36 to 18/36 of the same "bypass" diode string--It will not matter--The amount of power lost due to 100% shading will be the same.

    It sounds like you're saying that partial shading should result in only power loss of all cells in the same diode string. One thing I'm seeing is an entire panel string stops producing power basically after the bottom row of cells is in shadow -- so I'm not sure how to correlate that with your description.

    How can I tell where the bypass diodes are placed on a single panel, and how that placement relates from one panel to another?

    My panels are Phono Solar PS200M-24/F; Vmpp = 37.8V (12 rows x 6 columns = 72 cells)

    http://www.phonosolar.com/product_solar.php (Diamond / Onyx 190 - 210W Mono)

    Attachment not found.
  • BB.
    BB. Super Moderators, Administrators Posts: 33,431 admin
    Re: Rapid, large variations in panel output on cloudy days

    It depends on how those panels are wired... I believe I read that they put bypass diodes when there is a risk of >12 volts of reverse voltage... So, a 37 volt Vmp panel would have ~3-4 sets of bypass diodes. And the strings could be bypassed so you have "vertical rows" (left 2, middle 2, right 2), or they could be 4 with (3 bottom, 3 lower middle, 3 upper middle, 3 top)...

    If the shade covers all 3-4 bypassed strings (just a couple of cells in each string), then the whole solar panel output would be killed.

    This is not my field, so I really have know idea how it is normally done.

    Experimentally, you could just take a 1 foot wide strip of card board and try shading the bottom couple rows vs shading the side couple of rows and see what happens to your output.

    -Bill
    Near San Francisco California: 3.5kWatt Grid Tied Solar power system+small backup genset
  • AceNZ
    AceNZ Solar Expert Posts: 104 ✭✭✭✭
    Re: Rapid, large variations in panel output on cloudy days

    Sure, a few experiments would be easy, although I think the diode strings are probably vertical, since I know that shading the bottom row disables the entire associated string of panels.

    Which leads back to my original question. Is it common to suddenly lose all output from one or two out of three strings of panels when it's cloudy, even when the panels are all facing the same direction?

    Is there anything I can do to reduce the dropouts I'm seeing now?
  • boB
    boB Solar Expert Posts: 1,030 ✭✭✭✭
    Re: Rapid, large variations in panel output on cloudy days

    One extremely important question has not been asked or answered...

    What is your Voc of the strings and/or how many modules do you have in series for each string ??

    It appears that you have a 48V battery.

    If your strings, unshaded, normally park their Vmp just above your battery voltage, then the loss of
    one module may very well cause that entire string to go away when one module is shaded.

    boB
  • BB.
    BB. Super Moderators, Administrators Posts: 33,431 admin
    Re: Rapid, large variations in panel output on cloudy days

    Clouds should not cause output to fall to zero--Some light clouds, 50% reduction is common (and you may not even notice the reduction in light). And for deep dark storm clouds, you could see 90% reduction in power/current.

    Clouds, for a MPPT type controller (charging, GT Inverter, etc.) should not change the Vmp-array (as the array cools, Vmp-array should rise a bit).

    "Hard Shading" will usually reduce panel voltage--And can cut current/power flow to near zero (as discussed above)--Depending on the operational parameters of the attached controller (i.e., Vmp-array falls below battery voltage, below MPPT controller's operating voltage, etc.).

    If I understand your charts and question correctly--The second chart with light clouds varying power between ~2.5 and 5.0 kWatt--Nothing strange that I see.

    Our eyes operate on a logarithmic scale... Pretty much, our eyes can barely "see" a factor of 2 changing in lighting levels (i.e., if you were in a north facing room (for those on the other-side of the equator) for your home, you may not even notice the changing in available solar energy between 2.5 and 5.0 kW solar array output.

    Our eyes see a factor of 10x as a "major" change in lighting levels. You could use a light meter, a small solar cell connected to a DMM in current mode, or even, possibly a smart phone application to measure the differences in available light.

    Does your array controller let you read the V-array and I-array? I watch my GT inverter and see that the V-array voltage does not drift that much during the day (due to cell temperatures). While I-array varies explicitly with available sunlight.

    So, two things... If you do not see the factor of 2 changes in your graphs on a clear day--The hardware (panels and controllers) are probably working fine. If you only see these swings when there is some sort of clouds--I would not worry.

    If you can get your hands on a DC Current Clamp Meter/DMM, you could measure the current in each string (stable/sunny weather) and verify that every string current is identical to each other. If you find significant variations between strings (in identical solar irradation), there may be issues (panel problems, poor electrical connections, etc.).

    A secondary problem for GT Connected Inverters--There are times when variation/shutdown in GT power output caused by AC line problems. That can be high or low utility grid voltage (more common), or frequency out of spec, or power line trash/spikes etc. causing the GT inverter to time out for 5 minutes.

    -Bill
    Near San Francisco California: 3.5kWatt Grid Tied Solar power system+small backup genset
  • AceNZ
    AceNZ Solar Expert Posts: 104 ✭✭✭✭
    Re: Rapid, large variations in panel output on cloudy days
    boB wrote: »
    What is your Voc of the strings and/or how many modules do you have in series for each string ??

    Voc = 45.4V per panel.

    There are three strings of 10 panels each, so Voc per string = 454V.
    boB wrote: »
    It appears that you have a 48V battery.

    Yes. 8 x Trojan IND17-6V (925AH).

    DC from the panels goes to a Delta Solivia 5.0 MPPT inverter. AC from there goes both to the House Load and to a Victron Quattro 48/10000/140 inverter/charger, and from there to the batteries and/or the grid.

    I'm measuring the current drop at the AC coming out of the Delta.
    boB wrote: »
    If your strings, unshaded, normally park their Vmp just above your battery voltage, then the loss of one module may very well cause that entire string to go away when one module is shaded.

    You mean if the strings were feeding the batteries directly?
  • AceNZ
    AceNZ Solar Expert Posts: 104 ✭✭✭✭
    Re: Rapid, large variations in panel output on cloudy days
    BB. wrote: »
    Clouds should not cause output to fall to zero--Some light clouds, 50% reduction is common (and you may not even notice the reduction in light). And for deep dark storm clouds, you could see 90% reduction in power/current.

    I can understand a 50% reduction. I'm just surprised that it changes so quickly.
    BB. wrote: »
    If I understand your charts and question correctly--The second chart with light clouds varying power between ~2.5 and 5.0 kWatt--Nothing strange that I see.

    The range is more like ~1.5 to 5.0 kW (the dark green line).

    Here's a link to another version of the chart that only shows generation:

    http://pvoutput.org/intraday.jsp?id=24188&sid=22036&dt=20131227
    BB. wrote: »
    Does your array controller let you read the V-array and I-array? I watch my GT inverter and see that the V-array voltage does not drift that much during the day (due to cell temperatures). While I-array varies explicitly with available sunlight.

    Here's are readings from the GT inverter from a few minutes ago:

    AC power: 3706W
    AC voltage: 237V
    AC current: 15.8A
    AC freq: 50.0 Hz
    PV1 voltage: 332V
    PV1 current: 5.7A
    PV2 voltage: 338V
    PV2 current: 5.9A

    There are three strings, but apparently two of them are combined somehow before being fed into the GT inverter (single MPPT).

    It's a partly cloudy day today.
    BB. wrote: »
    So, two things... If you do not see the factor of 2 changes in your graphs on a clear day--The hardware (panels and controllers) are probably working fine. If you only see these swings when there is some sort of clouds--I would not worry.

    OK.
    BB. wrote: »
    If you can get your hands on a DC Current Clamp Meter/DMM, you could measure the current in each string (stable/sunny weather) and verify that every string current is identical to each other. If you find significant variations between strings (in identical solar irradation), there may be issues (panel problems, poor electrical connections, etc.).

    The clamps/meters I have only measure AC; I can see how measuring DC would be useful, too.
    BB. wrote: »
    A secondary problem for GT Connected Inverters--There are times when variation/shutdown in GT power output caused by AC line problems. That can be high or low utility grid voltage (more common), or frequency out of spec, or power line trash/spikes etc. causing the GT inverter to time out for 5 minutes.

    That's possible, but then I would expect GT inverter output to drop to zero, which it doesn't. The time measurement resolution of the Flukso current-measurement device I'm using is about 5 seconds at my desktop, 60 seconds at their website, and 5 minutes at PVOutput.
  • BB.
    BB. Super Moderators, Administrators Posts: 33,431 admin
    Re: Rapid, large variations in panel output on cloudy days

    Note there is no "storage" between solar panels and the GT Inverter input... The input to the GT inverter probably only stores a few 60 Hz AC cycles of power at most--So, a 5 minute per point plot (or even 5 second plot) is "forever" in those terms.

    The power output (and I-array current) is pretty much the instantaneous solar radiation value. So, yes, I would expect wide power swings in seconds to fractions of a second based on cloud activity.

    And you are correct, a 5 minute time-out due to hash on the AC mains would drop to zero--And most inverters would throw some sort of fault on the display (if you are watching for it).

    -Bill
    Near San Francisco California: 3.5kWatt Grid Tied Solar power system+small backup genset
  • AceNZ
    AceNZ Solar Expert Posts: 104 ✭✭✭✭
    Re: Rapid, large variations in panel output on cloudy days
    BB. wrote: »
    Note there is no "storage" between solar panels and the GT Inverter input... The input to the GT inverter probably only stores a few 60 Hz AC cycles of power at most--So, a 5 minute per point plot (or even 5 second plot) is "forever" in those terms.

    Yes, of course. The 5-minute and 60-second values are averages of once per second snapshot readings. The every 5-seconds number comes from a single reading.

    The Delta takes at least 60 seconds to recover from an AC sync failure, though, so I would expect to see dips like that on the Fluxso charts or with my desktop widget.
    BB. wrote: »
    The power output (and I-array current) is pretty much the instantaneous solar radiation value. So, yes, I would expect wide power swings in seconds to fractions of a second based on cloud activity.

    When I think of it in terms of the sun being covered or uncovered in a matter of seconds by a moving cloud, rather than a rising or setting sun, it makes more sense.
    BB. wrote: »
    And you are correct, a 5 minute time-out due to hash on the AC mains would drop to zero--And most inverters would throw some sort of fault on the display (if you are watching for it).

    Yes, I've watched the Delta drop out due a sync failure with the Victron (which raises its output frequency by 2 Hz when the batteries are full and the grid is offline), and it flashes all sorts of lights and messages.
  • solarix
    solarix Solar Expert Posts: 713 ✭✭
    Re: Rapid, large variations in panel output on cloudy days

    I've never heard of a Delta Soliva inverter, but MPPT strategies can vary greatly and some inverters do have difficulty achieving good efficiency during dynamic events like you are describing. I'd play around with a piece of cardboard as was suggested and see if you can characterize the problem more distinctly. Even though one panel's bypass diodes may work properly to maintain string current, the voltage is going to drop significantly which can be difficult for the inverter's MPPT to handle correctly.
  • AceNZ
    AceNZ Solar Expert Posts: 104 ✭✭✭✭
    Re: Rapid, large variations in panel output on cloudy days
    solarix wrote: »
    I've never heard of a Delta Soliva inverter

    Delta is based in Germany, Thailand and Taiwan.

    http://www.solar-inverter.com/au/en/296.htm

    Victron Energy is based in The Netherlands.

    http://www.victronenergy.com/inverters-chargers/quattro/

    Both are big in Europe, Australia and New Zealand.
    solarix wrote: »
    MPPT strategies can vary greatly and some inverters do have difficulty achieving good efficiency during dynamic events like you are describing

    They *claim* "Full power MPPT range" of 150 to 450V.
  • SolInvictus
    SolInvictus Solar Expert Posts: 138
    Re: Rapid, large variations in panel output on cloudy days

    Photovoltaic cells respond very quickly (microseconds, nanoseconds) to changes in incident light. To assess the order of magnitude, the shadow of a cloud moving at 20 miles/hour moves at 29.3 feet/second causing the shadow to cover a PV array in less than a second. Clouds can be thinner at their edges allowing more light through, and the sun subtends about .5 degrees in the sky which makes the edges of the shadows blurry. The edge of clouds also refract and diffract sunlight which can make bright and dim bands. When the shadow of a cloud passes over my off-grid PV panels, the readings on both my digital and analog current meters change quickly (in seconds). Readings taken at 1 second intervals ought to be able to record some of these effects for slow clouds, but not for fast ones.

    At your link, PV Output: Sunny Nelson Hybrid 6.000kW December 27, 2013 11:55 pm, I do not see any data nor link to data recorded at 1 second intervals. Are you saying that the high resolution readings show the power output decreasing from, say, 5 kW to 1.5 kW within one second?
  • AceNZ
    AceNZ Solar Expert Posts: 104 ✭✭✭✭
    Re: Rapid, large variations in panel output on cloudy days
    Photovoltaic cells respond very quickly (microseconds, nanoseconds) to changes in incident light. To assess the order of magnitude, the shadow of a cloud moving at 20 miles/hour moves at 29.3 feet/second causing the shadow to cover a PV array in less than a second. Clouds can be thinner at their edges allowing more light through, and the sun subtends about .5 degrees in the sky which makes the edges of the shadows blurry. The edge of clouds also refract and diffract sunlight which can make bright and dim bands. When the shadow of a cloud passes over my off-grid PV panels, the readings on both my digital and analog current meters change quickly (in seconds). Readings taken at 1 second intervals ought to be able to record some of these effects for slow clouds, but not for fast ones.

    All sounds reasonable.
    At your link, PV Output: Sunny Nelson Hybrid 6.000kW December 27, 2013 11:55 pm, I do not see any data nor link to data recorded at 1 second intervals. Are you saying that the high resolution readings show the power output decreasing from, say, 5 kW to 1.5 kW within one second?

    That link shows samples averaged over 5-minute periods, with some type of smoothing applied.

    The Hour charts at flukso.net show samples averaged over 5-minute periods, but without smoothing. I don't see a way to make those charts visible publicly, or I would (they also only retain the Hour chart data for 24 hours). On those charts, I've seen PV current drop from 5 kW to 1.5 kW within one sample period, and go back up again just as quickly.

    The highest resolution data I have, snapshot samples every 5 seconds, unfortunately doesn't have any charting or history, so I can't say how quickly the changes have happened there. It's overcast here today, but I'll try to capture some high-res data the next time it's partly cloudy.


    BTW, from what I can tell, "PV1" and "PV2" on my GT inverter (mentioned above) are some sort of internal measurements; the panels are only connected to one of the inverter's four DC inputs. The DC feeds from the three strings come together in a junction box with a breaker at the panels, and from there to another similar box right next to the inverter. I recall the installer saying something about combining them outside to save on wiring costs.