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

[ekarlson]

I was just looking at extending the PV array on my trailer to use three panels and was considering whether to wire the panels in series or parallel. I had seen all sort of posts making claims both ways - serial is better because it boosts voltage and allows charging under low-light conditions - parallel is better because shade on a panel takes out the entire serial array.

But as has been pointed out here, modern panels all have bypass diodes because, well, electrical engineers are clever and know that shade exists (I checked on 9 different common manufacturers and all of them had bypass diodes - in fact, some manufacturers have gone a step further with half-cut panels). These bypass diodes effectively remove strings of shaded cells from the panel which prevents them from dragging down power production. The other factor is that PV cells reach their nominal Voc with very little light - about 200 W/m2 from the literature I can find online - so even shaded cells quickly reach their nominal Voc which eliminates most of the concerns about low-voltage scenarios.

So I took my two new Renogy 100W panels, a Victron 100/20 MPPT and a depleted battery plus an 80W fan to act as a load. Wired everything up and then ran shading tests to see how much power the was actually delivered to the load under seven different shading conditions with the panels wired in series, and then repeated the tests for the panels wired in parallel. The reason for seven test cases for each configuration is that the bypass diodes impose a structure on the 36 PV cells in the panel, so it matters which combinations of cells are shaded. In the case of the Renogy panels, it was a bit odd because while the panel has 36 PV cells (arranged in 3 columns) there are only 2 bypass diodes and the diodes are not associated with the same number of PV cells. One bypasses 12 cells (Column A), and the other bypasses 24 cells (Columns B & C). Hence, the 7 test cases are as follows (note - shading was only applied to one panel in the array - the other panel was always left completely unshaded):

1. No Shade - should be 100% power
2. Shade on Column A - should be about 83% power (60 out of 72 cells)
3. Shade on Columns A & B - should be about 50% power (36 out of 72 cells)
4. Shade on Columns B & C - should be about 67% power (48 out of 72 cells)
5. Shade on Column C - should be about 67% power (48 out of 72 cells)
6. Shade on the bottom 4 rows (covers 3 cells in columns A, B & C) - should be about 50% power (36 out of 72 cells)
7. Fully Shaded (all cells, all columns) - should be about 50% power (36 out of 72 cells)

The tests were conducted at about 2PM local time, so close to maximal sun conditions. In each test case, I allowed time for the MPPT output to settle down, and ran the tests as quickly as I could to avoid varying sun and battery conditions.

What I found was that the shaded performance was about the same for both serial and parallel configurations in all 7 shade scenarios. This is what one would expect with the bypass diodes plus the fact that PV cells reach their nominal Voc levels with even small amounts of light. Since shade is low-light, not no-light, the shaded cells still seemed to reach their nominal Voc which prevents much of the power loss that a parallel array would see if any panel's output voltage were to drop below the others. Consequently in both configurations usable power correlated roughly with how many strings of cells were unshaded (as the bypass diodes effectively removed strings that were shaded) and therefore total current varies with the number of unshaded strings of cells but not Voc.

There is one scenario that I've not tested yet, which is to see how much power is actually available to the load over the course of an entire day. The interesting edge case is the low-light conditions seen and dusk and dawn - something < 200 W/m2. In principle, serial arrays would continue to generate enough voltage to turn on the MPPT charger where as the parallel array would not (the Victron MPPT manual points out that Vpv > Vbat + 5V in order to turn on the charger, and Vpv > Vbat + 1V to keep the charger turned on - so if the array is not generating sufficient voltage levels the power coming from the array will not actually make it to the load). However, this edge condition only occurs for a short period of time each day, and the panels are not generating that much power to begin with during those periods, so it is unclear to me whether this represents a significant factor in terms of watts-per-day that are delivered to the load.

I was going to post the measurements here, but when I repeated the tests with the panel order reversed, I got different results. This made me suspicious as it almost seemed like the bypass diodes in one of the panels were not working. So I popped open the wiring box on the panel that seemed to be misbehaving and, sure enough, it was missing one of its PV connections! See the attached picture - notice that the middle connector is not actually connected to the panel...

So, word to the wise, when buying new panels, check to see that the bypass diodes are actually wired correctly because apparently even well known companies like Renogy get it wrong sometimes...

Once I get the replacement panel, I will re-test and post the results. I think the results I saw when I was shading the non-defective panel are correct, but I want to ensure that I have two properly constructed panels before I post results.

 

[chrisski]

What I found was that the shaded performance was about the same for both serial and parallel configurations in all 7 shade scenarios

Odd. Completely different then when I tested it. Total output of three and four panels with A single shaded panel performed much better when in parallel, the bigger the shade, the bigger the difference.

I wonder if once your bypass diode is fixed what will happen to performance.

Parallel Panels and Shading versus Series Parallel

I tested some panels for output today when shaded. 2 Panels in Parallel (36 Cells Per Panel) -One panels with only 2 cells shaded 26% Reduction in Wattage -One Panel 50% shaded 54% Reduction in Wattage 3 Panels in Series (33 Cells Per Panel) -One panel Two cells Shaded 33%-50% Reduction in...

diysolarforum.com

 

[featherlite]

ekarlson,

I appreciate the work you are doing by posting the results of your testing.

Here is a photo of the diodes on my 100 watt 12v panel from Rich Solar. Notice that the middle metal tab is connected to nothing except the ends of the two diodes. Is that different from what is shown in your photo?

 

[timselectric]

That's typical.
They use the same junction box for all panels.
Whether it’s a 1 or 2 string panel.

 

[RoadTurtle]

Also consider the type of shading...

Where I had the 5th wheel parked, there was a thin pole casting a shadow across one of three panels in parallel. I lost the output from only that panel until the shadow cleared. In series, I'd expect to lose the entire string, or am I missing something?

I will very likely add one more panel to each string, and will rewire from 1s3p to 2s2p

 

[featherlite]

That's typical.
They use the same junction box for all panels.
Whether it’s a 2 or 3 string panel.

So, is there any defect in the wiring in the OP's photo?

 

[timselectric]

So, is there any defect in the wiring in the OP's photo?

Doubtful.
It's only a 100w panel. (Probably only 1 string)

 

[featherlite]

"Once I get the replacement panel, I will re-test and post the results."

I hope purchasing a replacement panel is not a waste of money.

 

[HighTechLab]

I think this discussion will find this article exceptionally interesting:

https://www.buyers-world.nl/pdfs/tigo-energy/TIGO-impedance-matching-eng-02-20-13.pdf

 

[ekarlson]

I am quite certainty that the panel is wired incorrectly because the other panel, also a Renogy RNG100D-SS, does have a connection on the center connector - see attached pic. Also, it makes no sense to use 2 diodes to create a single bypass - pointless increase in cost/complexity in constructing the panel.

 

[ekarlson]

Hm - actually, thinking about it more the serial results do not make sense to me now. Serial arrays work with matched current and aggregate voltage. But current is proportional to the number of illuminated cells. So in a panel with multiple strings separated by bypass diodes, when some cells are shaded, the panel is working off a reduced number of cells, which means a reduced current. That reduced current should impact the entire serial array... But of course if one of my panels effectively has just a single string, then shading the panel effectively takes the entire panel out of the array so the power drop would be equivalent to having a smaller number of panels rather than a reduced current across the entire array. This suggests that my test results were impacted by having that one panel wired differently than the others. It also suggests that the shade impact would be more significant is larger arrays (taking out 1 panel of 10 is a 10% reduction, but a current drop of 33% in one panel would drag down the entire serial array by about 33%). It would also suggest that when using panels that have multiple bypass strings that parallel would work better. Well, parallel would work better as long as the shaded cells are still producing something close to the nominal Voc. If the Voc of the shaded panel drops, that would be expected to impact the entire parallel array as it would drag down the Voc of the entire parallel array.

It will be interesting to see what happens when my return is processed and I get a new panel...

 

[ekarlson]

I was looking at the post that Will Prowse did showing the impact of shade on his serial arrays and I think I've realized where the issue lies and why I got apparently different results in my testing. My arrays are tiny - 2 panels. Mr Prowse's arrays were much larger - 10 panels. The short story is that for large arrays, given the electrical properties of PV cells, serial arrays where the panels have multiple bypass strings will be more severely impacted by partial shade than parallel arrays.

The logic works as follows:

1. When a panel is partially shaded, it loses amperage from the shaded bypass strings.
2. However, unless the shade is extremely dark the Voc of the shaded cells does not decrease significantly (according to the online literature that I can find, a PV cells has a near-nominal Voc at about 200 W/m2)
   1. I did confirm this with my panels. Fully illuminated I measured a Voc of about 20.5V. When I shaded one of the bypass strings, I still got a Voc of 19.1 V.
3. Hence if shade covers only some of the bypass string in a panel, the amperage is much more severely impacted than the Voc
4. For a parallel array, as long as the Voc remains roughly the same, the reduction in array power is proportional to the number of shaded strings vs the total number of cells in the array
5. For a serial array, having one element produce less amperage then the other elements results in loss of amperage for the entire array

So, for a 10-panel serial array - shading one string out of three in one panel reduces the amperage of the panel by about 33% but does not significantly impact the panel's Voc. Reducing amperage in one element of a serial array reduces the amperage for the entire serial array. So a 33% reduction in amperage in any panel with constant Voc reduces array amperage, and hence power, by about 33%.

For a 10-panel parallel array - losing one string out of three in one panel results in roughly a 5% reduction in array power. This is because the Voc for the partially shaded panel stays about the same, which maintains the Voc for the entire array (if the Voc of the shaded panel were to drop significantly, it would be a different story). Since amperage is roughly proportional to the number of illuminated cells, the amperage of the partially shaded panel drops by about 33%. However, the other 9 unshaded panel continue to operate at the same Voc and full amperage. Since amperage is roughly proportional to the total number of illuminated cells, in an array of 10 panels with 36 cells in each, we are getting amperage from 360 - 12 -> 348 cells out of 360, which is about a 5% drop in power, as long at the Voc doesn't change.

So it boils down to the size of the array and the fact that panel Voc does not drop off like amperage does when cells are shaded.

Note that if a panel has only a single bypass diode, then there is no such thing as a "partially shaded" panel. Any shade on the panel effectively bypasses the entire panel. In such a scenario the array simply performs as if there were fewer panels. Under those conditions, one would expect to see about the same power loss for the array with either serial or parallel arrays. Which means that for large serial arrays, partially shaded panels with multiple bypass strings actually impacts array power generation more severely than having just one bypass string because reducing a panel's amperage is more severe than completely bypassing the panel.

 

[Hedges]

For a serial array, having one element produce less amperage then the other elements results in loss of amperage for the entire array

Only if the MPPT is too stupid to try drawing more current, causing voltage to drop until bypass diode is activated.
If it is smart enough, you still get 9/10 panels producing, 90% power.
More likely, each panel has 3 diode bypassed sections, so if shade only covers part of one section you get 29/30 or 97% power.

10 in parallel, if one of 3 diode bypassed sections is shaded, voltage where that panel delivers full current is 2/3 as much and the other 9 panels keep voltage high. I expect 90% power.

Your SCC/MPPT may vary.

 

[featherlite]

ekarlson,

I also have a New Powa 12v 100 watt panel and it has the three tabs connected like your "non-defective" Renogy panels. See photo below.

However, I also have a third Rich Solar 12v 100 watt panel with diffeernt external dimensions and, like the first Rich Solar panel in the photo I posted above, also has only two tabs connected.

I am anxious to learn about the performance difference you find when you receive your replacement Renogy panel. I am not knowledgeable enough to know whether my Rich panels are substandard or not.

 

[ekarlson]

So the replacement panel arrived today, and I confirmed that the wiring is correct - 2 bypass diodes across 3 connectors with all connectors wired into the PV array.

the setup for the testing went as follows:

• Two Renogy RNG-100D-SS 100W panels, each with two fully wired bypass diodes
  • 36 PV cells arranged in 3 columns of 12 cells
  • As noted earlier I believe the association ov PV's to diodes is asymmetric - I think there are 12 PVs in the first column on the first diode and 24 PVs in the next two columns on the 2nd diode
• Victron Energy SmartSolar MPPT 100/20 charge controller
• BougeRV 3-to-1 MC-4 connector for parallel connection
• 10AWG MC-4 cables to connect from panels to MPPT
• 10AWG wires to connect from MPPT to battery/load
• Deep Cycle Lead Acid Battery with an 80W 12V fan connected in parallel for the load
  • Battery was depleted by letting it run the fan for 2 hours
• Power monitoring done using Victron's bluetooth app looking at the power that was delivered to the battery for charging
  • The MPPT was given about 15 to 20 seconds to settle down after each change of shade pattern
• Tests conducted at 3:30PM with the two panels arranged side-by-side facing the sun as directly as possible
• Shade was cast by lying the cardboard box that was used to ship the panels on the panel directly on top of the various cell groupings
• Shade was only applied to one of the two panels

The results in terms of parallel vs serial were "depends on which groups of cells are shaded":

Shade	Parallel	Series
None	183W	183W
Column A only - 12 cells	97W	150W
Columns A & B - 24 cells	96W	89W
Columns B & C - 24 cells	94W	121W
Column C only - 12 cells	99W	10W
Bottom 4 rows across all three columns - 12 cells	97W	88W
Fully covered - all cells, all columns - 36 cells	94W	89W

Some other interesting observations:

• The connector on the BougeRV 3-to-1 seemed a little flakey. On more than one occassion when I would first connect the panels the 3-to-1, the MPPT would only show about 12V coming from the panels. If I disconnected the 3-way and connected again, the voltage would come up to the expected 18V to 20V.
• Even with both panels fully covered, the MPPT still reported about 14V open-circuit coming from the panels, which would seem to confirm the graphs that I found online regarding Voc for a PV cell by irradiance - namely that it doesn't take much light at all for the Voc to rise to something close to the nominal Voc.

Some other resources that may be of interest to folks:
24-hour Solar Irradiance by Latitude and Day-Of-Year
PV Cell Voc by Irradiance

 

[octal_ip]

Great work with your testing, nice to see people putting in the effort to determine the real-world outcomes on a complete, working system.
So as with most big questions in life, the answer is "it depends", and each person would need to evaluate based on their own circumstances.

The only thing I would call out (as you already have) is that these results might scale very differently with larger numbers of panels in the mix.

 

[ekarlson]

I think the interesting observation is that shade on any part of one panel in the parallel array essentially removed that panel's contribution to the total power. But for serial arrays, the results can be better or worse than parallel depending on whether one is lucky or unlucky with respect to which groups of cells are covered. There is a catastrophic scenario where shade on an unfortunate grouping of cells appears to degrade the entire array's performance. From that perspective alone I would tend to go with a parallel array as it shaded performance is more predictable and does not have the disaster scenario found for serial arrays.

 

[Hedges]

What's going on with "Column C only - 12 cells" - 10W?

I must not have a good visualization of the setup, would have expected same as "Column A only - 12 cells" 150W

 

[ekarlson]

What's going on with "Column C only - 12 cells" - 10W?

I must not have a good visualization of the setup, would have expected same as "Column A only - 12 cells" 150W

I think it comes down to the fact that the distribution of cells to bypass diodes in uneven. That aberrant case happened when I shaded 12 of the cells in a string of 24. It does not work out the same as shading all 12 cells in a string of 12.

The following diagrams show what I believe to be the wiring of the Renogy bypass diodes and shows the difference between the shading for the 10W serial result and the 150W serial result:

 

[rfsam]

Just found this thread.

I believe series is better. If one string, so long as voltage remains within MPPT range.

Yep. I believe that too. Single string is the easiest and best for partial shading unless there are some limitations (SCC input Vmax, safety concerns and etc.). It puzzles me why people still think that parallel connection is better for partial shading conditions. Maybe that was the case years ago when panels didn't have bypass diodes? I see this being repeated in many Youtube videos by popular youtubers and etc.

If too many for max voltage, then series/parallel. I had 9s2p. I tried shading a single panel, and it appeared I only lost the output of a single panel, not two, because max power points of 8s and 9s are close enough in voltage.

Yeah. I tried to do some comparison doing 8s and 4s2p configuration with the same 12V 200w panels (8 panels total for the test). In 8s configuration one shaded panel was reducing output by 1/8, two - by 2/8 and etc... Bypass diodes were doing their job. But in 4s2p one shaded panel was knocking out the entire string, so I was loosing half of the power. Even when I tried shading 1 panel in each string power output was greatly reduced. I guess because of some V unbalance. So, I guess when you have 8 panels in the string (like what you had) the V unbalance is not that significant but what if two panels are shaded in the same string? I guess in this case unbalance will be high enough to block the entire string...

In my setup I have eleven 200w (12V) panels all connected in series to Midnight Classic 250 at the moment. Sometimes during the day I can get up to 3-4 shaded panels. That's just a space limitations I have. So I heavily rely on bypass diodes in my setup as the easiest solution and it does work. But recently I was thinking about TS4-A-O optimizers. I wonder if they will allow me to increase an output a bit (for single string config)? Also I'm a bit worried about bypass diodes "health". What would be your recommendation? I also cannot add more panels at the moment as I'm coming close to my SCC Vmax. One question about Optimizers bugs me: Can I do something like 6s2p config with individual Optimizers on each panel, so that they can re-balance V in each string? This way I could get same thing as what I get in a single string config - exlude only shaded panels, not just entire string if I have 2-3 panels shaded in either of the strings? I couldn't find an answer on Tigo's web site. I assume Optimizers are DC to DC converters, but do they have enough "intelligence" to do string balancing?


All my panels were purchased new in 2022 and 2023: BougeRV, HQST and Rich Solar. They are closely matched.

Thank you!