Adding a LF isolation transformer to a transformerless H5 grid-tie inverter(PV).

[Fflint]

I wasn't sure if this should go into the renewables section, but I guess it applies to inverters in general so I posted it here.

I have a modern grid-tie transformerless PV inverter in H5 configuration. In Europe almost all budget inverters that meet the current regulatory requirements do not provide isolation between the PV side and the grid. Those inverters, when they operate put significant negative and positive voltages on the PV input. This is mostly fine, but my PV panels require the negative electrode to be at zero to ground or positive (reasons below).  I've managed to find a low frequency medical isolation transformer at proper wattage to be able to isolate the inverter on the grid side and ground the PV-(minus) terminal hopefully.

However, before I risk my expensive(to me) inverter I decided to simulate everything in LTSpice and I discovered a potential issue I hope that perhaps someone could help me with(there are more issues, for example the inverter software measuring the grounding etc, I'm not worried about it at this stage - yet).

So below we have a simplified model of the inverter made up in LT-Spice(click to zoom in):
The isolation transformer can be disregarded in this schematic as AC neutral is tied to ground via 10 ohms here.


And here I show the waveforms present on the PV and AC output in normal operation. (click to zoom in)


You can hopefully see on the waveforms how the PV-(minus) terminal potential frequently reaches -600V. This is what I have to avoid.

So below we have a schematic of the same, but with isolation transformer "engaged" and PV- grounded.(click to zoom in)


And the respective waveforms. I added the red trace showing the differential voltage between AC live and "neutral" for clarity.(click to zoom in)


As you can see the goal for the PV-(minus) has been reached making it zero (or close to it) to ground. The differential AC voltage is still a nice sinusoid so the isolation transformer should make great use of it. However, the potential between AC (both live and neutral) and ground now contains very high frequency switching "stuff" and I'm not sure if this is not going to cause problems with my inverter. Does anyone have any advice how should I go about checking if this may be a problem? I can't open the inverter up due to warranty sticker, but I can probe various things from the outside. I have measuring equipment, rf generators, oscilloscopes etc available. I was thinking about starting with overcurrent protection inline with the PV and the isolation transformer in place, but putting low inductance potentiometers in both grounding paths(the AC neutral to ground and PV- to ground). Start with the AC side grounded and slowly shift to PV- side grounding hopefully catching any possible problems when they occur before they manage to blow up any parts. Any comments, suggestions?


The below is an explanation why I need this and alternative approaches It is relevant to this discussion only for context. I include it for information sake, but I would like to focus the discussion on the above.

So, why do my PV panels need it? Because they are thin film CIGS panels I bought for next to nothing at bankruptcy auction. They are made by a German company that went out of business because their panels suffer horrible potential induced degradation making them almost useless within few years if the minus terminal is at negative to ground potential for too long. Did I knew about it when I bought them?Yes, but I thought it is an opportunity to experiment. At worst I'll loose some time on my "hobby", at best I'll have a functioning PV system at 20% of the normal cost.
An alternative method to prevent PID is to glue conductive backing foil on the back of panels tied to each panel's PV- electrode therefore shifting "the ground" potential. The cause of PID is an electrostatic field in the soda lime glass used for the panel backing. The amount of PID(potential induced degradation) is directly proportional to the charge transferred by sodium ions in glass. If we can zero the field by making the back of the glass be at the same potential as PV- the problem is solved, ions become stationary again. I even bought enough one side conductive mylar foil to go down this route and it was my preferred one, but then I managed to get that isolation transformer for next to nothing. The inconvenience of having to glue that foil onto each panel, connect it reliably etc, set in and I decided to at least try the LF isolation transformer approach.

Hopefully this thread helps me in getting to the bottom of if I can use that isolation transformer and save myself the experience of gluing 30 sq. meters of mylar onto glass :-)

[moffy]

Could you clarify something? I assume that the PV input is made up of a number of series connected PV panels? If so then PV- for each panel will be different, or am I missing something?

[NiHaoMike]

The amount of high frequency noise on the AC output with respect to the inverter ground is going to be small thanks to conducted EMI requirements, but the DC input might have more since it only has to meet the radiated EMI requirements (as unlike the connection to the grid, it's not intended to connect to a variety of random equipment) which start very high compared to the switching frequency.

[Fflint]

Could you clarify something? I assume that the PV input is made up of a number of series connected PV panels? If so then PV- for each panel will be different, or am I missing something?

Hi, yes you are of course correct, so not all panels will experience PID(potential induced degradation) to the same extent. Pid is caused by the - being very negative to ground.
Those are 120V panels. 4 in series for a max voltage of 480V in my setup. So the one closest to the negative end will see the max negative voltage (-600-this is coming from the inverter), but the one on the positive end will at worst see only - 120V if not grounded. However if the minus rail of each string would be grounded the one on "the bottom" would see zero while every subsequent panel would see only positive voltage on its minus(that is fine).

The problem is that when I use the isolation transformer and ground the DC side I see lots of this scary "noise" on the AC side referenced to ground. I wonder if there is possibility thag it will cause me some problems.

The amount of high frequency noise on the AC output with respect to the inverter ground is going to be small thanks to conducted EMI requirements, but the DC input might have more since it only has to meet the radiated EMI requirements (as unlike the connection to the grid, it's not intended to connect to a variety of random equipment) which start very high compared to the switching frequency.

I guess it all depends if the circuit was built without any place where high frequency ~300V to ground could cause problems.

[NiHaoMike]

I guess it all depends if the circuit was built without any place where high frequency ~300V to ground could cause problems.

In the US, radiated EMI starts at 30MHz so the filtering just has to attenuate harmonics at 30MHz and above to an acceptable level.

That said, if the inverter has arc fault detection, most likely the filtering would be better so that the arc fault detection circuit won't have so much noise to deal with.

[moffy]

Thanks for the clarification. I expect that the inverter will have grounding internally at least for the metal work, but possibly elsewhere. That makes it a more complex issue.

[Fflint]

Thanks for the clarification. I expect that the inverter will have grounding internally at least for the metal work, but possibly elsewhere. That makes it a more complex issue.

Yes, definitely for metalwork. It has a huge passive radiator in the back. It is a sealed (IP65) unit that is supposed to be outside.

I don't think it has arc protection. The manual contains a list of possible errors and there is no "arc fault" or similar. Perhaps I'll have to do the gluing.

[moffy]

From the looks of your sim the grounding of the solar panels could increase the insulation stress on the power devices (I am assuming grounded heatsinks), because it looks like it applies a DC shift relative to ground, providing a possible failure point.

[Phoenix]

Hi Fflint

Your simulated PWM generation for the H5 topology appears to act as a simple unraveller circuit. In practice it's a bit more clever in a way to reduce parasitic coupling during the zero voltage PWM state. The AC current only ever free wheels around S1/D2 and S3/D5 which is decoupled from the DC bus by S5 in your diagram.

You can check out this paper for the pattern:
https://www.mdpi.com/1996-1073/11/11/2912/htm

As for your specific question - I wouldn't expect any problem in the power electronics from your proposed scheme. Normally the grid is "stiff" in a common mode sense and any HF stuff appears inside the inverter. With your proposal the inverter bus would be common mode stiff and as you have found the AC output will have the HF. The EMI filter inside the unit and isolation transformer will go a long way to stopping a lot of this from getting into the grid.

As others have said the internal insulation stresses may be higher than the original designed values. You can check this out in simulation once you have corrected the PWM by probing the switch source WRT ground with/without the isolation transformer. Even if they are slightly more stressed I wouldn't be concerned - it won't be much.

The biggest hurdle I would see is the firmware. It will likely have mains earth detection, array isolation/insulation resistance detection, RCD - lots of things to make it safe in it's non-isolated condition. You will probably need to kill all of these functions!

Finally - this will no longer be compliant to any safety standards and you will take the risks. Whether anyone polices this or not is another matter.

[Fflint]

Hi Fflint

Your simulated PWM generation for the H5 topology appears to act as a simple unraveller circuit. In practice it's a bit more clever in a way to reduce parasitic coupling during the zero voltage PWM state. The AC current only ever free wheels around S1/D2 and S3/D5 which is decoupled from the DC bus by S5 in your diagram.

You can check out this paper for the pattern:
https://www.mdpi.com/1996-1073/11/11/2912/htm

As for your specific question - I wouldn't expect any problem in the power electronics from your proposed scheme. Normally the grid is "stiff" in a common mode sense and any HF stuff appears inside the inverter. With your proposal the inverter bus would be common mode stiff and as you have found the AC output will have the HF. The EMI filter inside the unit and isolation transformer will go a long way to stopping a lot of this from getting into the grid.

As others have said the internal insulation stresses may be higher than the original designed values. You can check this out in simulation once you have corrected the PWM by probing the switch source WRT ground with/without the isolation transformer. Even if they are slightly more stressed I wouldn't be concerned - it won't be much.

The biggest hurdle I would see is the firmware. It will likely have mains earth detection, array isolation/insulation resistance detection, RCD - lots of things to make it safe in it's non-isolated condition. You will probably need to kill all of these functions!

Finally - this will no longer be compliant to any safety standards and you will take the risks. Whether anyone polices this or not is another matter.

Thank you for the link and the reply. I haven't seen this particular article.

The inverter has all required fault detection built in. This is a list of "protection features"


I originally intended to find a way of bypassing many of them in "isolated" mode later, but the longer I think about it the more I start to consider the alternative solution (with the foil). It would be lots of work to reverse engineer the firmware and modify it and physical modification would break every possible code as you said.

I did try the foil backing on one one panel yesterday. I used 3m 77 spray adhesive and 3 separate single conductor very thin AWG 30 wires (10in or so each simply placed between the panel and foil) for the connection. Using multiple wires will allow me to periodically monitor connection quality. Measuring right after the installation I only get few ohms of resistance. Those "panels" are frame less - two sheets of glass glued together type and they are held to my (self made) ground mounting stand with (manufacturer made) clamps that contain a big chunk of rubber that holds the glass. Also the one-side conductive foil is mylar backed (50 microns). So the insulation should be good enough for over 2kV (they will never go above 500V).

Two major worries I have about the foil solution now are: the glue stops working after few months of rain and frost and the edges when they become unglued will carry up to 360V to ground potential.

So to mitigate I'll connect the backing foil to each panel's minus via a multi MegaOhm resistor so there is no shock risk even if the foil unglues itself. Also I'm thinking about perhaps painting over the edge of the foil with good outdoor varnish (marine spar varnish) to help seal it, but this is probably beyond the scope of this thread.

[moffy]

there are conductive paints that might help:
1. Galvanising paint, its meant for outdoors
2. Nickle spray on paint, really nice but expensive.

[fourtytwo42]

You could always buy a s/h transformer inverter, I have a Fronius IG15 and there are several in that series, very good. Personally I wouldn't touch the transformer-less stuff with a bargepole!

[Fflint]

You could always buy a s/h transformer inverter, I have a Fronius IG15 and there are several in that series, very good. Personally I wouldn't touch the transformer-less stuff with a bargepole!

I spent a good amount of time searching for a small(3kW), single phase, transformer based inverter that meets current EU regulations and doesn't cost 5x the money.

I found nothing.

To illustrate the problem let me say this about the inverter you mentioned. The latest EU certificate mentioned in the paperwork is from 2008. This is ancient stuff. For anything grid connected we need 2016 NC Rfg compliance with later changes.

By now I've glued the mylar foil to all my 20 panels. It was lots of work (half an hour per panel).

I hope the foil doesn't restrict the cooling too much. I'll find out in the summer I guess.