DISCUSSION: PV Series vs. Parallel. Is parallel REALLY better for partial shading situations?

[Hedges]

3 Panels in Series (33 Cells Per Panel)
-One panel Two cells Shaded 33%-50% Reduction in Wattage
-One Panel 50% Shaded 69% Reduction in Wattage
-One Panel 70% Shaded 88% Reduction in Wattage

Something has to be very wrong in that setup. Either,

1) no bypass diodes (I see these are flexible panels)

2) Shaded panel put out less current, and SCC did not lower voltage to where the other two panels could push current past for only 33% reduction (regardless of how much that one panel was shaded.)

Screenshot:
45.92V --> 28V
3.5A --> 3.0A

Voltage dipped below 30.6V (2/3), and current dipped.
This does not appear to be activating a bypass diode.
If you had Y-cables to tap, you could check intermediate voltage.

 

[chrisski]

I feel the results I posted above were accurate compared to what I see with my flat portable panels with bypass diodes, I just don’t have any data written down. I set my portable panels up a dozen times a year and I partially shade them and watch the dips in output to make sure they are setup correctly.

I do think that if shading and SCC max voltage is not a factor, series is better than parallel wiring losses from high amperage, but with what I’ve measured, any real shading requires as many in parallel as possible.

Something has to be very wrong in that setup. Either,

1) no bypass diodes (I see these are flexible panels)

I have not popped the covers on either of the flexible panels or glass panels, but both have cases that should contain diodes. In a couple of weeks, I’ll be on top of the Rv roof and check then.

2) Shaded panel put out less current, and SCC did not lower voltage to where the other two panels could push current past for only 33% reduction (regardless of how much that one panel was shaded.)

Screenshot:
45.92V --> 28V
3.5A --> 3.0A

Voltage dipped below 30.6V (2/3), and current dipped.
This does not appear to be activating a bypass diode.
If you had Y-cables to tap, you could check intermediate voltage.

Y cables are a bit hard to get to, but may try to measure with a multimeter at some point.

 

[SignatureSolarJames]

OK, lets say you have 8 panels vs 2 strings of 4

-the bypas dioode is for the 3 internal strings of cells in each panel (or 6 if you have split cell). Bypass diodes are over 12 years into adoptionn cycle, so definitely not something new.
If one panel of 8 is making 200w out of 460 because of shade the others will always be limited to 200w.

I once had a classic sunnyboy and one string of 12 panels, shade was cutting production by 30%, I went back and modified to 2x6 and the production went up 15%

 

[Hedges]

If one panel of 8 is making 200w out of 460 because of shade the others will always be limited to 200w.

200/460 = 43% of full power

7/8 = 88% of full power

If MPPT functions well, there is a bypass diode, and 7s Vmp is enough, that should operate at 88% not 43%.
That would be up by 45% (of full power) or up 100% (depending on what you compare to.)

An increase of 15% *might* be due to improved efficiency at lower voltage (closer to AC Vpeak after whatever transformer) but Sunny Boys according to their published curves only roll off a few percent.

Doesn't mean it didn't happen, but seems like something not working right. Some bad diodes would be one possibility.

We had a guy come pitch optimizers to our manufacturing company almost 20 years ago. That's when I ran the experiment of shading one panel out of 9s2p (Sharp 165W polysilicon into SWR 2500U). There were varying clouds so it was not a well controlled experiment, but reduction in power was about 1/18th, not 2/18th or 9/18th.

MPPT getting stuck is certainly a thing, but not expected with your 12s. Every few minutes Sunny Boy will do a sweep to see if lower voltage gets full current, rather than partial current from what light hit that panel. Not real frequent so would take patience. My 9s2p, following curve would find shifted peak, no need for a sweep. Half-cut panels, which are more like (20s2p3s)12s (if 120 half-cells and array is 12 panels) would have multiple peaks.

Maybe today I'll run more experiments. But it is overcast.

 

[efficientPV]

I operate in extreme shade with most of the day 70% of panels are shaded. Some panels only get an hour of sun a day. I use 2S6P which will provide sufficient power for my needs. Parallel shaded panels still provide some power. One upside is that even in overcast and rain the array still provides enough power to provide basic needs.

 

[SignatureSolarJames]

200/460 = 43% of full power

7/8 = 88% of full power

If MPPT functions well, there is a bypass diode, and 7s Vmp is enough, that should operate at 88% not 43%.
That would be up by 45% (of full power) or up 100% (depending on what you compare to.)

An increase of 15% *might* be due to improved efficiency at lower voltage (closer to AC Vpeak after whatever transformer) but Sunny Boys according to their published curves only roll off a few percent.

Doesn't mean it didn't happen, but seems like something not working right. Some bad diodes would be one possibility.

We had a guy come pitch optimizers to our manufacturing company almost 20 years ago. That's when I ran the experiment of shading one panel out of 9s2p (Sharp 165W polysilicon into SWR 2500U). There were varying clouds so it was not a well controlled experiment, but reduction in power was about 1/18th, not 2/18th or 9/18th.

MPPT getting stuck is certainly a thing, but not expected with your 12s. Every few minutes Sunny Boy will do a sweep to see if lower voltage gets full current, rather than partial current from what light hit that panel. Not real frequent so would take patience. My 9s2p, following curve would find shifted peak, no need for a sweep. Half-cut panels, which are more like (20s2p3s)12s (if 120 half-cells and array is 12 panels) would have multiple peaks.

Maybe today I'll run more experiments. But it is overcast.

How is it possible for a series of panels to output more per panel than the least productive panel?

 

[sunshine_eggo]

How is it possible for a series of panels to output more per panel than the least productive panel?

Umm... this is what bypass diodes are for.

Your statement is only true for current, not power.

 

[timselectric]

How is it possible for a series of panels to output more per panel than the least productive panel?

Because the voltage is still additive. And the current stays constant.
In a parallel situation the voltage would be decreased for all. To an equalized voltage.

 

[Hedges]

How is it possible for a series of panels to output more per panel than the least productive panel?

Umm... this is what bypass diodes are for.

Your statement is only true for current, not power.

If 10 panels in series produce 10A and one in series produces only 5A, you can either have 10 x Vmp x 5A, or 9 x Vmp x 10A.
The 9 push 10A through the 1's bypass diode. But only if SCC draws voltage down further.

I think SCC has capacitor on input, then a switched inductor, then capacitor on output. Switching the inductor acts as a controllable current source. By pulling more amps from input capacitor, voltage from PV string is pulled down, activating bypass diode.

Because the voltage is still additive. And the current stays constant.
In a parallel situation the voltage would be decreased for all. To an equalized voltage.

With parallel panels (or strings), Voc of a shaded panel is similar to Vmp of illuminated ones, so voltage doesn't decrease. (even in total darkness it draws some current when driven to Vmp, but considerably less than Imp or Isc.)

SCC will pull voltage lower only if that provides higher power, reducing power from illuminated ones by moving off peak point, and increasing current and power of partially shaded one. Several in parallel, I expect it to ignore a single shaded one.

 

[Nobodybusiness]

All this math is giving me a headache. I'm just going to cut that damn tree down when my wife is gone to Walmart.

Simple solutions!!!

 

[octal_ip]

I'm probably a bit late to the party, however I did a somewhat related test last year, and found parallel is a more optimal arrangement for when whole panels are facing towards/away from the sun: https://diysolarforum.com/threads/blocking-diode-question-please.51248/post-653783
This would be similar to shading as the panel facing away from the sun had no direct illumination at all, just ambient light.

For easy numbers, let's say this panel is a 300W 60 cell with 36Voc/30Vmp and 10A Imp.

Thus this panel will drop to 24Voc/20Vmp and 10A Imp when the cell is shaded.

What happens if this is in parallel with another identical panel @ 30Vmp/10A?

Well, like batteries, panels have to be at the same voltage when in parallel, so the only way you'll get ANYTHING out of the shaded panel is if the array is below 24V with optimal at 20V.

Assuming max power, the moment the cell is shaded, power drops from 600W to 300W.

The MPPT tries to pull more current and finds that it can get max output of 400W at 20Vmp. Yes, this will actually be a LITTLE higher than 400W because the unshaded panel will output a little more than Imp, but I'm ignoring that since I don't know what the number will be, but certainly somewhere between Imp and Isc.

I understand that you've simplified this for the sake of the example, however it's important to remember that a shaded cell won't reduce its Voc to 0v. In practice the Voc change during shading can be surprisingly small - the change in ability to source current though, is huge.

While in operation, the load (MPPT) would be pulling the panel/cell voltage down well below their Voc to Vmpp, so the minor Voc and Vmpp reduction from the shaded panel/cell won't be enough to make much of a difference to the operation of the parallel cells/panels. Certainly not as much as in the example where the shaded cells/panels are considered dead and producing no voltage at all.

If you performed a test where you completely covered one of the cells/panels, allowing practically no light through at all, your example would be closer to reality. However in practice shading (unless cast by very nearby objects) is usually not enough to completely disable a panel/cell. It'll alter their Voc and Vmpp a bit, but not nearly as much as their ability to produce current. I believe it's for this reason that parallel usually performs better than series when shading is applied.

 

[Hedges]

However in practice shading (unless cast by very nearby objects) is usually not enough to completely disable a panel/cell. It'll alter their Voc and Vmpp a bit, but not nearly as much as their ability to produce current. I believe it's for this reason that parallel usually performs better than series when shading is applied.

That indirect light doesn't mean the series string has less power available. Just that if SCC doesn't drop voltage low enough, it will harvest only power produced at that string's low current. In other words, a firmware shortcoming.

 

[zanydroid]

If you performed a test where you completely covered one of the cells/panels, allowing practically no light through at all, your example would be closer to reality. However in practice shading (unless cast by very nearby objects) is usually not enough to completely disable a panel/cell. It'll alter their Voc and Vmpp a bit, but not nearly as much as their ability to produce current. I believe it's for this reason that parallel usually performs better than series when shading is applied.

But in the case of a series string, why wouldn't the MPPT sweep detect that the lower operating voltage has higher output power, and force the operating voltage low enough to bypass the 50% shaded 1/3 of a panel? Suppose there is a single solar module.

Run at V = V * 0.5 * Impp = 0.5 * V * Impp
Run at 2/3 V = 2/3 * V * 1 * Impp = 2/3 * V * Impp

disregard ~1-2 volt voltage handwaving, I'm keeping it simple and leaving out the differences in voltage from activating bypass and the partial string shading.

 

[zanydroid]

However in practice shading (unless cast by very nearby objects) is usually not enough to completely disable a panel/cell

Suppose you have a 10p string. Suppose you have 0.1 I_mpp due to shading on 1/3 of one module. Then with parallel, that module can only generate V * 0.1 * I_mpp. Because most likely it is better to operate the parallel string at V, since 2/3V (which will allow that bypass to activate) would be a net negative when the string is that wide.

Suppose you have a 10s string. Then MPPT can trivially drop 9 2/3V and only lose a tiny amount of the production from that part shaded module. Basically the 30W on that 1/3 of the module. Which you actually probably can never retrieve for this shading pattern even with MLPE, because those operate at module level. Great!

Let's say that part shaded module is 300W. Then in the parallel case you get 30W from it. In the series case you get 200W from it.

This does not constitute a proof in all cases, I still want to see some kind of sampling or exhaustive based simulation of the space.

The MLPE would have value if you have consistent shading across the whole module. With bypass diodes the whole module production is lost. With MLPE you can recover the full production.

 

[time2roll]

Shade sucks no matter how the panels are wired. I assume it is possible to cherry pick the shade scenario where either could show better or worse.

 

[Will Prowse]

I actually tested this years back on my RV roof when I had shading issues in some parking spots, and parallel configuration produced more energy over the course of a day than if I put them all in series. Great discussion above and I figured series would be better, but parallel was actually superior. There was always a noticable increase.

It does depend on the wiring configuration, SCC being used, how fast the mppt can track and change when shading does occur (some are very slow), nominal voltage of battery bank and how well it's matched to the panels, and the converters input to output differential.

Victrons excel with shading because they can track multiple power points and track very quickly.

But overall, I found parallel did produce more. It was quite interesting. I always figured the bypass diode would only drop the voltage a little bit, but it seems to really throw something off and reduce output.

 

[zanydroid]

If parallel was amazing in a lot of cases, surely we would see more MPPTs optimized for parallel operation? As things stand, the vast majority of MPPTs that can boost are in microinverters. If parallel was so good, wouldn't we see MPPTs that incorporate boost so that they can charge 48V batteries etc?

(Throwing out a strawman. I appreciate that series has a lot of other design advantages)

 

[Will Prowse]

If parallel was amazing in a lot of cases, surely we would see more MPPTs optimized for parallel operation? As things stand, the vast majority of MPPTs that can boost are in microinverters. If parallel was so good, wouldn't we see MPPTs that incorporate boost so that they can charge 48V batteries etc?

(Throwing out a strawman. I appreciate that series has a lot of other design advantages)

Gosh it really depends on the components and voltages being used. In low voltage with a 12V battery and 20VOC panels, parallel does offer an advantage. Especially if the shading is covering a few panels. If all the panels were in series and feeding a higher voltage battery or the grid, the string voltage may decrease to a point that the mppt can not operate. The voltage drop across bypass diode and shaded cells can drop the voltage significantly.

 

[timselectric]

I think that's the situation.
At lower voltage, parallel is better. But at higher voltages, a small loss in series is negligible.

 

[zanydroid]

As things stand, the vast majority of MPPTs that can boost are in microinverters.

And while I'm taking a hit off this pipe, why can't we have MLPE that are microinverters with the inverter, grid tie, and anti-islanding chunk deleted.

Parallel me the 340VDC rail directly onto the branch circuit as a voltage source with no weird logic nonsense.

Keep some RSD functionality.

 

[octal_ip]

I actually tested this years back on my RV roof when I had shading issues in some parking spots, and parallel configuration produced more energy over the course of a day than if I put them all in series. Great discussion above and I figured series would be better, but parallel was actually superior. There was always a noticable increase.

It does depend on the wiring configuration, SCC being used, how fast the mppt can track and change when shading does occur (some are very slow), nominal voltage of battery bank and how well it's matched to the panels, and the converters input to output differential.

I'm the same - I've done the tests both on the ground and on the roof, and at least for my systems, parallel produced more power by quite a large margin.
While it's an interesting thought experiment, there's too many variables at play to isolate a single factor and draw conclusions around the behaviour of an entire system, particularly when there's not linear or binary relationships between these variables and conditions constantly vary.

 

[Shimmy]

I understand that you've simplified this for the sake of the example, however it's important to remember that a shaded cell won't reduce its Voc to 0v. In practice the Voc change during shading can be surprisingly small - the change in ability to source current though, is huge.

This is what I see as well-- most of my shading does not activate the bypass diode as the voltage drop is minimal. System configuration really has to be looked at holistically.

 

[Hedges]

Shade sucks no matter how the panels are wired. I assume it is possible to cherry pick the shade scenario where either could show better or worse.

The goal is to have a system where shaded panel(s) or potions of a panel cause no more loss than the diode-bypassed portion.
What we don't want is for one shaded panel to prevent others from contributing.

Optimizers might do this, if they can boost sufficiently. Microinverters should do this, with the reduced efficiency of having to boost voltage so much (about 6x). I think MPPT per series string does it (except if you're trying to capture power from a panel with 50% reduced current.)

 

[Hedges]

I re-ran my old test of shading one panel in a series/parallel array.

"SunPower" branded Sunny Boy 8000US
SunPower SPR-327NE-WHT-D panels, 8s2p array, 5232W STC

First data with unshaded panel was out of family from the others, probably more clouds.
The two tests with "B" string having one panel shaded had very similar results, which I averaged.
The unshaded test which I used (didn't discard) occurred between those two.

Would like shading one panel out of 16 to produce 93.8% as much power, but it was 90.8%; a 3% penalty for both strings being a bit off their Vmp.
The string with one panel shaded ideally would have been at 7/8 of Vmp for an 8-panel string, 87.5%, but it was at 90.8% of that voltage. The unshaded string would ideally have been at 100% of voltage, but was also at 90.8%. (The power vs. voltage curve falls off steeper on the right, so Vmp for the two parallel strings is closer to Vmp of a 7s string.)

Operating a bit above its Vmp, the string with a shaded panel delivered 83.1% of Imp. Operating 9.2% below Vmp, the unshaded string delivered 97.3% of Imp.

	"A" Amps	"B" Amps	Vmp	Power
Unshaded	4.11​	4.44​	392​	3722​	<-- discard (low light?)
"B" Shaded	5.36​	4.42​	356​	3687​
Unshaded	5.53​	5.44​	392​	4059​
"B" Shaded	5.40​	4.62​	356​	3683​
"B" average	5.38​	4.52​	356.00​	3685.00​
"B"/unshaded	97.3%​	83.1%​	90.8%​	90.8%​
		7/8		15/16
		87.5%​		93.8%​

It seems we each reconfirm what we found worked best.
Some arrays and shading conditions drive the outcome, but also MPPT algorithm of the SCC.

 

[Hedges]

Is that a vent pipe poking through a panel on the right side?

... or are you just happy to see the sun? ?