Switching between parallel and series connections (solar)

[solvejskovlund]

TL;DR Summary

How can a solar array be reconfigured (parallel vs series connections) based on sunshine conditions. Will relays do the job? How to avoid that a failing relay cases a dangerous short circuit?

We're off grid at 57degrees north. Our source of electricity is solar panels, with a diesel generator as backup. The solar has served us well, until this November, where we had almost 8 weeks of 0 sunshine. I got sick of running the generator. It's noisy, it needs refueling, smells... So I started experimenting with the solar system. What i found was that if I connect all the panels to a small solar charge controller that I had laying around as a spare, we got enough charging to run the equipment we need on a daily basis, even during during days of snowfall!
When the clouds go away, this small spare charger gets WAY overloaded. Hence, I need to find a way to use the main charger when the sun is up, and the spare charger when clouds block the sun. The problem is that the spare controller requires the solar panels to be connected in parallel short strings (max 3 in series), while the main controller requires long strings (at least 6 in series, 9 to be able to handle peak power). Currently the switchover is performed by me getting up on the roof, freezing my fingers of while disconnecting/reconnecting a bunch of connectors. I need a better solution.

I don't know why this spare controller does so much better than the main during cloudy days. One factor is that it has way less power consumption. Another might be that the Max Power Point Tracking algorithm is better.

First I was thinking to find an other spare charger, but with the ability to use the long strings. I've only been able to find one that might work. It is very expensive (11x what I paid for the current spare), has to be shipped from US to Europe, expensive shipping and tax, expensive returns, and there is no data available that can tell if it will be able to replace the current spare controller during cloudy days. Going this route will be an expensive experiment. I don't feel like going that route.

My second option is to make a network of relays behind the solar panels that makes me able to use a control signal to "reconnect" all panels with the flick of a switch. So I started drawing a connection diagram. What comes to my mind is that if all relays are controlled by the same signal, there is a risk that one relay switches before another, causing reverse voltage over a string of solar panels. This might damage the panels. I can avoid this if I split the control signal so that one group of relay gets the time to switch before the other gets the switching signal. But what if one relay gets stuck? Is there a way to implement some kind of feedback into this circuit? Something that can tell the second set of relays that one of the first relays has failed? Maybe relay is not the correct thing to use for this purpose? (Operating temperature is -30C to +30C, it will be in sealed box behind the solar panels (in shade).)

Maybe there are better ways to do this? Off course I'd like this electronic to consume near 0 power while in "clouds" state.

The attached drawing shows how I'm thinking 9 panels can be converted from 3 parallel strings of 3 (relaxed state, no signal), to a single string of 9 panels. The upper drawing shows how panels and control box are connected, while the lower drawing shows the inside of the control box. Drawing shows the relays in relaxed state (no input signal). When the switch signal input receives power, the upper relay will disconnect a panel string + from the "Bus+" (Charge controller +) and connect to the lower relay. The lower relay will disconnect a panel string from "Bus-" and connect to the upper relay. The sum of the two relays is that the negative of panel x will be connected to the positive of panel x+1. When signal power goes out, the relays will flip back to relaxed state.

The risk here is that if both relays in one control box goes from relaxed to engaged state before any of the relays in the other control box, there will be a condition where a string of 6 panels is parallel connected to a string of 3 panels. I imagine this will damage the 3. And the same risk will apply when signal power goes out.

So far all my ideas for relay feedback circuits has been a failure.

Any ideas for better ways to do this?

 

[berkeman]

Are you familiar with MPPT converters?

https://www.solar-electric.com/learning-center/mppt-solar-charge-controllers.html/

 

[rbelli1]

Are you familiar with MPPT converters?

Max Power Point Tracking algorithm is better.

 

[berkeman]

LOL, so it was mentioned somewhere in that long post? I read it over like 2-3 times and didn't see it mentioned so that's why I asked.

If the OP understands MPPT, why in the world is there a question about parallel/series/relays/etc?

 

[rbelli1]

They know what they need but are looking for a more cost effective alternative to use the parts at hand.

BoB

 

[anorlunda]

TL;DR Summary: How can a solar array be reconfigured (parallel vs series connections) based on sunshine conditions. Will relays do the job? How to avoid that a failing relay cases a dangerous short circuit?

When the clouds go away, this small spare charger gets WAY overloaded.

I agree with @berkeman , MPPT is the most likely solution. It sounds like your "spare" MPPT controller is undersized. That suggests a higher capacity MPPT controller as the one and only controller might be your best solution.

By the way, the instruction sheets for MPPT controllers specify the series/parallel combinations to use. They are not all the same.

 

[solvejskovlund]

If there was a way to evaluate the MPPT algorithms under different conditions before buying, there wouldn't be a problem buying the most suitable one. Problem is that such information is not available to any product. You have to buy them, wait for the correct weather, and test. As weather changes constantly, these products has to be tested side by side.

Both my controllers are MPPT. And they are from the same manufacturer. The spare is the older. So in that sense, the spare should be worse, not better, than the main. But I guess when they design the products they aim for max yearly production. Production during clouds does not make any noticeable difference in those numbers, hence they ignore those conditions.

But when you live somewhere where clouds can be thick for 8 weeks straight, those numbers turns out to be the most important of all. How much power that is produced during a sunny day is way more than enough, no matter how badly the algorithm is implemented.

 

[solvejskovlund]

That suggests a higher capacity MPPT controller as the one and only controller might be your best solution.

A bigger controller will consume more power. Running my two controllers side by side during cloudy day showed: the main controller with 12 panels CONSUMED 18w from the battery. The spare controller with 6 panels (same angle as the 12) CHARGED 21w to the battery. I could buy a bunch of controllers to test. It will be expensive, it will be time consuming. I may be able to have a result by December next year. Or I could make some control circuits to remotely do what I now do manually, that I know works, and it will cost less than one tenth of a new controller. A circuit like that shouldn't be hard to create, nor expensive. And it will give results faster than the experiment.

Btw, the experiment with several controllers will fail even before it starts. The reason is that there aren't really any stand alone solar charge controllers to chose from that supports more than 150V solar arrays. And out of the 7 hybrid solar charge controllers I ordered that were able to run with solar array of less than 150V, 7 was canceled by the sellers because they didn't produce them any more. I had to opt for 500V system. And I doubt there is such a thing as a hybrid solar charge controller that has low power consumption (even with the inverter part switched off).

To put it easy, the solar charge controller I would like the have does not exist. But it can easily be made from the controllers I already have by using some relays and some sensor logic to secure the system.

Now, please stop messing about changing controllers, unless you actually know of a product that can do the job, that is still in production and can be delivered to Europe. If you do know of such, please post link. If you don't, please stay on topic for this forum thread.

 

[rbelli1]

Now, please stop messing about changing controllers,

That may be the only solution so don't get mad at the suggestion.

What is the peak power and voltage? Relays and contactors get real expensive real fast for DC operation.

BoB

 

[Rive]

I had to opt for 500V system.

That settles it, then. That is totally not good for relays.
TBH I don't think you should touch it at all, considering the regulations and dangers of high voltage DC.

BTW the relays would eat up far more than you can gain.

Maybe a battery & autostart system would do better.

 

[Rive]

I got back to a keyboard, finally.

The thing is, you would need to design for the worst case, which is ~ a false switching at summer, broad daylight. On the other hand, the only gain you get is at winter, and is really minuscule. The disparity is just too big, especially if you aim to comply with all the regulations (and even if you aim for that, as a DIY stuff you can't get proper documents for the thing). So - just does not worth it. Keep it nice and clean.

If possible, then maybe you can get a different (high voltage, low power) controller instead of the oversized one for winter usage.

Also an option to change the system so that the generator would only charge the battery pack. To build an automatic system to start/stop the generator in need is low(er) voltage stuff, completely doable. And so you won't get that running all day.

 

[anorlunda]

I may be able to have a result by December next year.

Running my two controllers side by side during cloudy day showed: the main controller with 12 panels CONSUMED 18w from the battery. The spare controller with 6 panels (same angle as the 12) CHARGED 21w to the battery.

If those are big panels, say 400w each, it sounds like you are working to optimize the production of 20w from a 4800w system during extreme periods of low light. Is that the case?

 

[sophiecentaur]

the main controller with 12 panels CONSUMED 18w from the battery.

Now, please stop messing about changing controllers,

I could point out that you may not be able to find the 'equivalent ' performance of a relay based system under a range of 'marginal conditions'. The number of possible series / parallel relay arrangements would be limited and, away from optimum, they could do more harm than good.

I guess the best argument for or against the relay method would be to look at existing systems and compare the number of each that are in common use. All over the world, the fashion has been to move from mechanical to solid state systems. Would your requirement really call for a different philosophy?

 

[Baluncore]

I see two ways of doing the switching, both require the PV array be reduced to 3 modules, making 6 wires from the PV array to a control box.

1. The series/parallel switching could be done with five poles of two position switch. There would be flyback diodes to protect the switch contacts. You would need to flip the switch or switches by hand to change between series and parallel modes.

2. An automatic system would not use relays, but would employ MOSFETs and diodes, to do the switching. It would be more of a challenge to implement that solution in the field because it requires custom designed electronics and so has safety and legal implications.

When switching DC current sources, to prevent high-voltage inductive voltage spikes appearing during the changing configuration, the elements of the array can be individually short-circuited, (by say an isolated MOSFET). That is a little understood technique and must be applied with care.

Zero volts times full current represents zero power lost in the switch or the current source. The short circuit must not be removed until there is a load or a snubber to conduct the current.

 

[jrmichler]

The problem is that the spare controller requires the solar panels to be connected in parallel short strings (max 3 in series), while the main controller requires long strings (at least 6 in series

Can you wire your 12 panels into two modules, with each module wired 3S2P (two parallel strings of three panels in series)? Then connecting the two modules in parallel would result in 3S4P for the total system, while wiring them in series would result in 6S2P for the total system. One DPDT switch that is rated for the total voltage and current can switch between the configurations. That switch must be break before make, and ideally would have a center OFF position. Such switches are used as manual transfer switches for backup generators.

You might want a second such switch to transfer between the two controllers. If you never, ever forget to disconnect the controller before switching the panels, then there would be no current through the panel switch during switching so no switching transients.

 

[Baluncore]

Three equal sub-strings of PV array are connected by three DPDT switches. Each sub-string Vpv is controlled by one DPDT switch. All switches should be set to select the same mode.

Turn off the HV inverter before switching.

This arrangement prevents HV being applied to the LV inverter. The low-power LV inverter is always connected, unless the spare switch contacts are used to isolate it.

 

[solvejskovlund]

If those are big panels, say 400w each, it sounds like you are working to optimize the production of 20w from a 4800w system during extreme periods of low light. Is that the case?

That sums it up quite well. All though panels are 270w each, and I do get quite a bit of production even on the most cloudy days when they are connected in "low light mode". I'm surprised that the panels are charging more than 100w, even they are half covered in snow and the clouds are so think that when I look up at the sky at 12:00 and thinking that if I were at an unknown place without a compass nor gps, there would be no way I could tell which direction is south! This discovery is so valuable to me that I can not let it pass.

 

[solvejskovlund]

I guess the best argument for or against the relay method would be to look at existing systems and compare the number of each that are in common use. All over the world, the fashion has been to move from mechanical to solid state systems. Would your requirement really call for a different philosophy?

I've never designed a solid state circuit. I am aware they exist, and my understanding is that their advantage is when switching is required to happen fast or frequent. My requirement is that switching will happen at most twice a day, and it's ok if it takes a couple of minutes. I also have the impression that SSR has more power loss than a traditional relay, which is ok on the sunny days, but not on the cloudy days. That's why, at the initial drawing, I designed all relays to be in the relaxed state (no power supplied) on cloudy days.

 

[solvejskovlund]

When switching DC current sources, to prevent high-voltage inductive voltage spikes appearing during the changing configuration, the elements of the array can be individually short-circuited, (by say an isolated MOSFET). That is a little understood technique and must be applied with care.

Zero volts times full current represents zero power lost in the switch or the current source. The short circuit must not be removed until there is a load or a snubber to conduct the current.

I think the short circuit during switching is not needed because there will be 0 current for quite a while after switching. The switching process will be:
1) disconnect solar array from charger 1
2) engage switch signal
3) connect solar array to charger 2

At the current time, I'm looking into automating step 2 only. Currently this involves me getting into safety harness, climb the roof, freeze my fingers of while fiddling with the connectors, and climb back down. At some point I might look into automating all the steps, but for now I'm happy to keep my fingers warm.

The risk will off course be that if, for some reason the switch signal is lost while charging is in progress. If the system is designed so that 0 signal means that the circuit reminds in its current state, that is not a problem. Is the circuit is designed so that it jumps to parallel mode (cloudy day mode) when signal is lost, the problem should be minimal as the main charger would switch off when that happens (it has a build in relay (at least some part that says 'click') that is triggered by high voltage on the solar input). An unintended switch from serial to parallel will make the voltage drop enough for this relay to disconnect. All though it does take some time for this to react.

An unintended switch from parallel to serial connection will for sure be dramatic.

My biggest concern with the relay approach is that one relay reminds (either slow responding or stuck) in "parallel mode" while the others switch to "serial mode". And as this may experience temperatures from -30C to +50C, mechanical parts, like a relay, are more likely to fail.

 

[solvejskovlund]

Can you wire your 12 panels into two modules, with each module wired 3S2P (two parallel strings of three panels in series)? Then connecting the two modules in parallel would result in 3S4P for the total system, while wiring them in series would result in 6S2P for the total system. One DPDT switch that is rated for the total voltage and current can switch between the configurations. That switch must be break before make, and ideally would have a center OFF position. Such switches are used as manual transfer switches for backup generators.

It is possible, yes. All though there are 18 panels (for comparing chargers side by side I wired 12 to main charger and 6 to spare charger), so 3 modules then. They could be wired in groups of 3s2p, switching between 3(3s2p)s=9s2p and 3(3s2p)p=3s6p.
Unfortunately my roof and wall space does not go in a multiple of 3, so that would either require some extra cable streches, or more switches. Both approaches doable.

 

[anorlunda]

All though panels are 270w each, and I do get quite a bit of production even on the most cloudy days when they are connected in "low light mode".

That makes your questions easier to understand. The manufacturers of MPPT controllers might be neglecting that operating regime, leading to your performance problems. Some of our answers may also have been off-target for the same reason. Most people give up on solar in those low-light conditions. My neighbor has a big solar PV system. I asked him how much he gets in the winter months, he said, "Zero." That's not true, but that's how he views it so he shuts it all down.

p.s. Children who grow up near the artic circle never learn about the Sun rising in the East and setting in the West. At mid-summer it sets in the south (if at all) and mid-winter it rises/sets in the south.

You'll see a lot of caution about using relays to switch DC currents. That's likely to lead to trouble. But if you guarantee zero current when things are switched, it's not a problem. Switching AC is less of a problem because AC current passes through zero 60 times per second.

So pay attention to the details of the suggestions you already got in this thread. Those guys know what they're talking about.

 

[solvejskovlund]

Three equal sub-strings of PV array are connected by three DPDT switches. Each sub-string Vpv is controlled by one DPDT switch. All switches should be set to select the same mode.

Turn off the HV inverter before switching.

This arrangement prevents HV being applied to the LV inverter. The low-power LV inverter is always connected, un<less the spare switch contacts are used to isolate it.

Actually, that way of connecting would be quite fail safe. The HV charger has a relay build in that is triggered by solar input voltage. Once voltage is in the LV range, it will disconnect. And when switching from LV to HV, the relay has a delay for connecting.
But the LV charger has no such relay. And thats where the high currents go. I theory the LV solar current can reach (spike) to 54A, before the limiter kicks in. If sw1 is the last to be switched from HV to LV it will have quite a current to deal with.
Currently I use 2 pole DC auto fuse breakers to (dis)connect the chargers.

(wonder what the MPPT algorithms would say about that setup with both chargers connected in HV mode.)

 

[solvejskovlund]

Switching AC is less of a problem because AC current passes through zero 60 times per second.

So pay attention to the details of the suggestions you already got in this thread. Those guys know what they're talking about.

Actually the (grid tied) AC current passes through zero 120 times per second in the US. 100 times here. (Just paying attention to details, as you said.)
Considering your name, you're living not far from me. I believe your neighbor has the exact same situation as me during November this year. But he is grid tied, could flip a switch and still have a working household. I had to start the noisy generator, and realize that I had no cable ready for connecting it. We had not had the need for the generator since the solar and batteries was installed during the fall last year. We had no need for low light solar power until that day either. At first I was thinking we must accept to run the generator once in a while. But when we were heading into the 7th week of zero sunlight, I go so fed up that something had to be done. Wind turbine.... there were almost no wind during those weeks. Waterpower from a near by creek... doable, but 300m away, across a forest owned by a quite unfriendly farmer. We'd have to hide the cable quite well. (The creek is own by a very friendly farmer.) My experiments with the spare solar charger I'd call a big success. It produces more electricity than we need even through the darkest days. (As long as weather is cold, and the outside fridge/freezer doesn't need power.)

 

[solvejskovlund]

View attachment 319003

I like this a lot because of it's quite failsafe without any sensors. A simple modification is to replace the DPDT switches with some DPDT relay or mosfet or SSR. How could this failsafeness be kept when putting two of them in parallel? Even if there is such a thing as a 4 pole relay, it will be unpractical to have all poles in the same spot. I'll need them distributed over a 10-15m stretch. (Or run long cables.) My panels are in 2 areas, 10m cable between them.

 

[sophiecentaur]

My requirement is that switching will happen at most twice a day, and it's ok if it takes a couple of minutes.

I wasn't suggesting you should SSRs but I was suggesting re-considering an intelligent solid state MMP controller.
Do you have any evidence of the amount of possible output power that your series / parallel system is actually achieving / wasting (i.e. actual efficiency)? A switching system will certainly get some current when none would be available with a parallel connection - so that's good. Your easily measured 18W of loss with a controller might be significantly less than with the simple two state solution but you may not know it.
I do agree that sticking with what you know has advantages but you are operating beyond the envelope so a clever system could be a good solution. The downside of changing would be long development time and loads of measurements, adjustments or trying a range of expensive controllers.

 

[solvejskovlund]

I keep my eyes open for other controllers, but there simply is no standalone 200v+ solar charge controllers on the market within a reasonable price. The only one I found cost tripple of my main charger, inverter, spare charger, fuses, cables... I don't want spend that much to test a controller that does not give any data that tells if it may solve my problem or not, without any return option!
The two data (power consumption in minimal light, and minimum Voc for charging) needed to evaluate if a charger will fit my needs are never given. They are never tested for during third party tests either.

Today, the day before the shortest day of the year, we had thick clouds and fog. Visibility of about 100m. I even had to use a headlamp to refill our animals water in the middle of the day. That has never happened before.
Open circuit cell voltage was 0,446v at 11.30am. The small controller managed to charge 62wh during the day, with only 8 panels (4s2p) connected. (I realized that the other 4s2p circuit was switched off when I went to read the results at the end of the day.)
It's not much power, but still impressive that it actually did charge anything in such conditions. The main charger will not charge at all if Cell Voc is less than 0,58v. This shows how major the difference is between the two controllers.
On a typical day with thick clouds the cell Voc is around 0,53v. Today was the first day since I installed the spare controller that the battery has not been recharged to the same level as the day before.

It doesn't really make sense to talk about power loss in a mppt charger. The charger that happens to have the best mppt algorithm for the test conditions may have the biggest loss (lowest efficiency) and still be the one that provides most charging power to the battery. The only important thing is how much charge it puts into the battery during the test.

 

[sophiecentaur]

It's not much power, but still impressive that it actually did charge anything in such conditions.

Yes. Very impressive and it would seem to be something you could rely on in winter. You'd always be able to supply a small, essential load.

It doesn't really make sense to talk about power loss in a mppt charger. The charger that happens to have the best mppt algorithm for the test conditions may have the biggest loss (lowest efficiency) and still be the one that provides most charging power to the battery. The only important thing is how much charge it puts into the battery during the test.

I used the term efficiency in response to your 18W loss figure. If your test conditions are the most representative then I'd agree. You've clearly put a lot of effort and thought into this and have a working system so I can hardly tell you what you should be doing. The only thing is that switch mode charge control could give the best possible use of available VI with low loss. But your application would probably call for quite a lot of specialist knowledge and a lot of trial and error.

My only brush with MPPT controllers was in the context of my sailing cruiser and, to be honest, the cost of the most sophisticated units was not really justified in my case. A larger battery bank would have removed the need for anything other than charging from the alternator in my case.

 

[solvejskovlund]

Thought you were supposed to sail, not run the engine.
You know there are props that can be submerged for charging while sailing? They also work at anchor if there are currents.

The 18w is what it pulls from the battery when the charger claims to be charging at very low rate. It consumes more power once there are more solar power. I have not collected numbers for that. The MPPT makes it hard to compare because for numbers jumping up and down a lot. When reading the amps, by the time my eyes has moved to the volts, the amps has changed significantly. I could take a picture of the meters to get reading of both at the same time. Even include the BMS in the picture, but then there are various delays in all displays as well, and they average over different intervals.

The good thing about the small charge is that the battery lasts longer. Actually, with the charging rates I've seen in cloudy days, a battery that lasts 40 hours would be sufficient. That makes the system way cheaper, and saves a lot of weight in case of a boat/camper installation.

I'll place an order for DPDT relays (there was non available locally) and try the circuit Baluncore suggested. It might do magic.

 

[sophiecentaur]

Thought you were supposed to sail, not run the engine.

HaHa. Yes- in principle but choose a random day and a random destination and there will be at least an hour's worth of motoring (whilst hanging one's head in shame.)

The MPPT makes it hard to compare because for numbers jumping up and down a lot

Agreed. 'Tests' would need to involve a lot of data and data-logging. Pretty tiresome if you don't actually enjoy that bit of life. ;-)

 

[Baluncore]

I have now taken a look at what might happen with long inductive cables to the switch box.
When any PV module is disconnected while current is flowing, there will be an inductive voltage spike that will arc over and burn the switch contacts. With DC the arc may continue and not be extinguished by voltage reversal as is normal with AC. To prevent that arc, a snubber circuit is required, as shown in the diagram. The snubbers use no power in normal operation.

Each component of the snubber has a specific purpose.
When module x is switched out, the positive spike charges Cx through Dx, limiting dv/dt.
When another module is switched out, the reverse voltage spike is caught by D1x.
After any switch event, Rx = 1k, restores the idle voltage across the capacitor Cx.

Diodes should be fast, and rated at least for momentary current of the PV array current. Heat sinks will not be required by chunky diodes, as the switch event is fast and rare.

Capacitor selection for say a 20 amp module current, 50 volt over voltage, and 10 us duration, will require;
C = 20 amp * 10 usec / 50 volts = 4 uF minimum.
The capacitor voltage rating will need to be at least 50 volts above the maximum PV array voltage.

I would consider scrounging the high voltage electrolytic capacitors and diodes out of old switching computer power supplies.

 

[Tom.G]

I would consider scrounging the high voltage electrolytic capacitors and diodes out of old switching computer power supplies.

I would recommend using Non-Inductive, Metallized FIlm capacitors rather than old Electrolytics.

Electrolytics fail with age by drying out, especially at elevated temperatures. They also have poor high frequency response, which they will see during switching transients.

Metal FIlm do not have that aging problem, they self-heal if they develop an internal short circuit, and the non-inductive ones have decent high frequency response to suppress the switching transients.

Cheers,
Tom

 

[Baluncore]

I would recommend using Non-Inductive, Metallized FIlm capacitors rather than old Electrolytics.

This is not running above 10 kHz in a switching supply. It may only be needed a couple of times each day.
The snubber does not need fast capacitors, it needs to ramp up voltage while current ramps down, and so prevent sparks. There is plenty of time and voltage headroom during the transient.
An electrolytic has distributed inductance and capacitance, a bit like a transmission line. If the inductance of the electrolytic was high, a parallel 10 nF ceramic could be connected in parallel, but that is what the first turn of foil in an electrolytic actually does.

Electrolytics fail with age by drying out, especially at elevated temperatures.

The capacitors in this snubber idle with no ripple current, so they will run cold.

I am advocating a cheap experimental solution that will work OK and do the job, rather than an expensive overkill on a military budget. Gold plating the electrodes and polishing electrons is not needed either.