Shading, Serial vs. Parallel, this time with data

[Ryban]

Background:
There are many discussions on this forum and elsewhere about the merits of serial vs. parallel solar panels when it comes to mitigating effects of partial shading (e.g. boom, rigging, wind vane, etc etc).

Probably for this scenario the very best thing is to have every panel on its own controller -- but lets ignore that for now since often it is impractical, and it potentially has its own issues at low voltages/cloudy days.

Otherwise, most people seem to suggest that putting panels in parallel yields the best protection from the effects of one panel becoming partially shaded. This thinking is generally a result of assuming that increased resistance on the partially shaded cell drops the throughput of the unshaded panel. However, most modern panels have a bypass diode across the entire panel, or the individual strings of cells on a panel, depending on the size of the panel. This diode is triggered on a reverse bias, which can occur when the MPPT charger picks a current higher than would otherwise flow through the shaded cell. This triggers the reverse bias on the diode, and current flows through it, instead of the string of cells.

In parallel operation, the bypass diodes do not come into play. A shaded string on any one panel in the parallel array, will not necessarily affect the power output from the other panels in the array, depending on the MPPT controller. Theoretically, it should still choose a target voltage that maximizes power coming out of the unshaded arrays, but this voltage will no longer be ideal for the partially shaded array, and its output will be reduced.

Honestly, I couldn't sort out all the contrasting opinions, each with seemingly sound arguments, and I also couldn't find good testing data -- maybe I didn't search hard enough. I got fed up and decided to just test it myself.

The test
I did a simple test using a single Victron 100/30 MPPT controller, and 2x 50W Newpowa panels, 2x 100W Renogy panels. The 50W panels have a single bypass diode (1 string), and the Renogy panels have 2 bypass diodes (2 strings).

The panels were either put in parallel or series for the tests. I did not mix the 50W and 100W panels at any point.

The load was bulk charging a 280AH LFP battery.

For most of the test I completely covered 1 cell on one of the panels, and alternated between which cell I was covering to make sure there weren't any variations. I also did a "pole" shadow test by taking a 1.5" wide piece of wood and laying it across the strings on one panel.

I repeated these tests over the course of an hour, switching between parallel and serial configurations. Environmental conditions ranged from cloudy, partly cloudy, to full sun in this time frame.

Results

For single-cell shading, series produces more power. For the 100W panels in series, the panels produced 22% more power than the same panels in parallel. The 50W panels, unsurprisingly with only a single diode, were about equal in either configuration.

For stick shading across all strings, parallel produced more power, by about 15%. However, the sample size is pretty small for my stick-shading testing compared to the single-cell shading tests.

So, series or parallel?

If only small portions/single strings are expected to be shaded, series seems like the clear winner in my testing. But more practically, the type of shading we see tends to be more similar to the "stick" testing I performed, and in that case, the parallel configuration had a leg up.

I also think other factors may affect the results of these tests, including how many panels are in series/parallel and how many of them are suffering from partial shading, what panels are being deployed (how big are they, how many bypass didoes do they have), and the type of MPPT controller in use.

If I have time this week, I may perform the tests again and focus on full-panel shading, differing panel angles to the sun, and other factors that are commonly affecting our arrays.

[fxykty]

Awesome testing, thanks!

Another test is when one panel of a pair is 1/3 or 1/2 shaded, such as by the boom or a sail. On our cat that’s a common scenario with our bimini panels. Our panels are series wired in pairs either side of centreline. Each set of two panels is on the same side of the boat and when sailing the panel closer to the centreline can be partially shaded. They perform OK, but would they be better if wired in parallel?

[Mike OReilly]

Great test Ryban . Actual data is always preferred.

Certainly on my boat, the realistic situation is having all, or most, of a panel shaded. This is why I run all panels in parallel. It's good to know your test seems to support this approach, but I look forward to future data.

Not to diminish your efforts, but much like anchor testing, there are so many variables, many of which influence the others, that it is hard to generalize. Boats are different, shading is different, panels are different. But having real-world data can help guide our choices.

[Ryban]

Originally Posted by fxykty

nother test is when one panel of a pair is 1/3 or 1/2 shaded, such as by the boom or a sail.

This is a great question and actually the reason I did the testing, as I'm trying to figure out how I should wire up the panels on our dodger that will be under the boom (see photo). To that end, I did some more "realistic" testing today with some interesting results below.

Originally Posted by Mike OReilly

Not to diminish your efforts, but much like anchor testing, there are so many variables, many of which influence the others, that it is hard to generalize.

This is very true, and I suspect the important differences are actually the panels being used and the charge controller algorithm. I'm really doing this testing for myself though, as these are the panels that will be going on our dodger, and we'll be using a similar MPPT controller for them.

Updated results
Today was a cloudy/rainy day, but I managed to have enough light through the clouds to get some testing done at lunch.

Today I focused on 3 things:

• Differing angles (one panel gets less light than the other)
• Soft boom-like shadows across or along the strings of cells
• Hard shadows, that completely wipe out a string

The results are more or less consistent with the previous testing (refer to attached plot).

Differing angles
I had one panel oriented towards the sun, and the other oriented toward the sky 90-degrees opposite. In this configuration, I'm using the parallel case as the baseline since it produced more power. Relative to that, the series configuration produced about 25% less power. This is a situation where the bypass diodes are not being utilized since the reduced light on the panel is not sufficient enough to trigger any reverse biasing of the diode.

Soft boom-like shadow
For this test I used a 6" diameter tube (maybe it was less), and laid it either across a single string of cells, or diagonally across the panel. Orientation did not seem to matter too much. In this test, the series configuration produced on average, 16% less power.

Hard shadow that covers entire string
In this test I used a piece of wood that completely covered a string of cells. This may or may not be similar to a boom shadow. Even in a boom shadow scenario, the cells still have some visibility to the sky, so I suspect this is not the most realistic test.

In any case, in this test, the parallel array produces 18% less power than the series configuration.

This test triggers the bypass diodes, and it's apparent as the MPPT controller searches for the best power point, as it will drop to 0W before coming back up.

Conclusions

It seems that, in my limited testing with my specific hardware, that parallel is a little better in dealing with the types of shadowing/angle impacts that would affect output from solar arrays on a boat.

That said, in situations where shading is extreme enough to trigger bypass diodes, series is better. I'm not sure if bypass diodes would be triggered more easily with more panels in a string, or on panels with more strings (they shouldn't, right?), but most of the shadows I see on my boat would not be sufficient enough to trigger the bypass diode.

[Uplander]

Good stuff. Thanks for data.

[fxykty]

Another decision regarding parallel vs series is how much current will be running through the wires to the controller? With large >300W panels the voltages are plenty high for MPPT controllers and battery voltage differential for early mornings and late afternoons, but the currents can start adding up in parallel configurations.

[s/v Jedi]

All these tests confirm my data. It is important because there are many in the forum claiming otherwise as the result of an old YouTube video that somehow was deeply flawed (one cell shaded killed the whole series array and many keep repeating that here).

Aiming one panel 90 degrees different than another is extreme and not realistic but it supports the general rule of thumb:

When you physically mount multiple panels in an array that appears like one big panel (all together and in the same pane) and this array has no boom or other object crossing over it then connect the panels in series.

When there is a boom going across then divide the array into two arrays, one on each side of the boom, and connect the panels that are on the same side in series and each side to it’s own controller.

[barcoMeCasa]

slight drift but sort of relevant, for anyone running signalk/openplotter on a raspberry pi there's been an app for a while which reads the bluetooth data from victron smartshunt/smartsolar etc. Used to be a complete pain to get the MAC Address & Advertisement Key to set it up but now the android victron app displays these so it's quick to set up.
https://community.victronenergy.com/...-protocol.html
So you can save everything your solar regulators & shunt are up to with the write to database app to really see deep down what's going on.
Though the internal bluetooth adaptor in the Pi is not the strongest so needs turned off & on again once in a while, usb/bluetooth adaptor would prob solve that. Just ask on the openplotter forums or discord for help to speed up the setup if interested.
https://forum.openmarine.net/
https://discord.gg/5wVvuGmt

thread drift over, back to where we were then....

[s/v Jedi]

Originally Posted by barcoMeCasa

slight drift but sort of relevant, for anyone running signalk/openplotter on a raspberry pi there's been an app for a while which reads the bluetooth data from victron smartshunt/smartsolar etc. Used to be a complete pain to get the MAC Address & Advertisement Key to set it up but now the android victron app displays these so it's quick to set up.
https://community.victronenergy.com/...-protocol.html
So you can save everything your solar regulators & shunt are up to with the write to database app to really see deep down what's going on.
Though the internal bluetooth adaptor in the Pi is not the strongest so needs turned off & on again once in a while, usb/bluetooth adaptor would prob solve that. Just ask on the openplotter forums or discord for help to speed up the setup if interested.
https://forum.openmarine.net/
https://discord.gg/5wVvuGmt

thread drift over, back to where we were then....

You don’t need to do that because a GX device will log all that to the web portal

[flightlead404]

Thanks for the testing.

As far as "shading" goes, I rarely see shading that actually mimics the shading we see on our boats. Or at least, that I see on my boat. As in your pictures, the shading simulation is done my placing an object directly on, or very close to the panel. This leads to a very sharply defined area with very dark shading. That doesn't really happen. Instead we see shadows from further away objects, like masts that are much lighter and less well defined, but may cover more area.

I wonder what difference that would make on your testing.

On my boat I found a way to place the panels aft of the book where there will be little to no shading other than from the mast, which is still reduced, or when the sun is at low angle and therefore not really producing power anyway.

I found better power production in series, especially on cloudy days when getting the voltage higher was important.

[barcoMeCasa]

Originally Posted by s/v Jedi

You don’t need to do that because a GX device will log all that to the web portal

Not everyone has a GX device though, the victron ble app could be very useful for anyone running signalk on a Pi already, don't have to do anything but it's free if anyone wants it.

Grafana is a quite a beast to get started on but immensely powerful for anyone into any kind of in depth data analysis, pretty much industry standard for time based data.

And probably of no interest to 99.9% on here

Back to the thread..

[s/v Jedi]

Originally Posted by barcoMeCasa

Not everyone has a GX device though, the victron ble app could be very useful for anyone running signalk on a Pi already, don't have to do anything but it's free if anyone wants it.

Grafana is a quite a beast to get started on but immensely powerful for anyone into any kind of in depth data analysis, pretty much industry standard for time based data.

And probably of no interest to 99.9% on here

Back to the thread..

But when you have the Pi anyway, you can run it as a GX device… nothing to buy

[Ryban]

Originally Posted by barcoMeCasa

So you can save everything your solar regulators & shunt are up to with the write to database app to really see deep down what's going on.
Though the internal bluetooth adaptor in the Pi is not the strongest so needs turned off & on again once in a while, usb/bluetooth adaptor would prob solve that. Just ask on the openplotter forums or discord for help to speed up the setup if interested.

Victron also provides a VenusOS image for free that you can load onto a raspberry pi and basically get the functionality of CerboGX, which does all this as Jedi mentioned. I use VE.Direct-to-USB cables and feed all the data into the Pi this way, rather than bluetooth. I have the VenusOS send all its data to a SignalK server I have running elsewhere, but the VenusOS image also comes with a SignalK server you can enable to run on the same device.

On the Pi I have running SignalK, I also have InfluxDB running and Grafana. I can then collect all this data very nicely that way.

Originally Posted by flightlead404

As far as "shading" goes, I rarely see shading that actually mimics the shading we see on our boats. Or at least, that I see on my boat. As in your pictures, the shading simulation is done my placing an object directly on, or very close to the panel. This leads to a very sharply defined area with very dark shading. That doesn't really happen. Instead we see shadows from further away objects, like masts that are much lighter and less well defined, but may cover more area.

....

I found better power production in series, especially on cloudy days when getting the voltage higher was important.

Although I held the objects to the surface of the panels in my tests, the "tube" and "stick" test did not completely block any cells, and allowed some current to still flow through these cells. As a result, these tests did not trigger the bypass diodes for those strings, and this is why, in those tests, the parallel configuration performed better. I think this is similar to the effect soft shadows would have on a boat. As long as the cells are getting enough power from the ambient light, and the bypass diode isn't triggered, then the parallel configuration will perform better.

In contrast, when even one cell was completely covered, the bypass diode triggered, and the series configuration performed better. Again, I don't think this happens frequently on a boat, unless your laying rags down on your panels.

I think your second point is good though and worth testing. I suspect the series panels would perform better on cloudy days, all else being the same.

[Pete7]

This is our previous 300w set up before switching to 600w. I spent some time one day watching the shadows on the panel with the yacht tied to a pontoon and facing due south. Although the back stay was very close to the panel, it didn't cause any visible shadow. The mast did but didn't completely block a whole cell instead parts of a number of cells. Also it moved quite a bit even in a calm marina. What was annoying is the shadow from the yacht mast next door early in the morning, then ours, finally late afternoon the shadow from the yacht on the other side of the pontoon.

It might be that you find the stern 100w panels need to be in parallel and those under the boom series or individual panel to MPPT.

Pete

[sailcrazy]

Following. Our 4x80w panels, mounted on the aft davit, are old, destined to be replaced, and the installation is an issue I'm trying to figure out.