SMA SunnyIsland + Enphase microinverters + A/C compressor load = enough electrical noise to interfere with microinverters?

[daklein]

Shucks, still happening. It worked fine with the A/C on for several hours about 11a to 1p, but then later from about 3-5p it's not working.

Today the event logs for the microinverters reveal a couple things of interest.

During the afternoon not working time period, the grid-tie micros (connected in beyond the basement panel with the A/C) show all three events, grid instability, ACFreqOOR, and ACVOOR on both L1 L2. The 'off-grid' solar micros connected in at the closer garage panel show only grid instability, ACFreqOOR.

During the working pretty normal period, both show GridInst and ACFreqOOR, but no ACVOOR. I think these ACFreqOOR events are from the SMAs limiting the solar output, normal behavior with the FLA near full. The grid-tie system shows only one or two ACFreqOOR events; it's freq trip limit is set to 61.3 vs. the default 60.5 of the 'off-grid' micros. The original goal there is to keep the GT system always producing because there's an 11c/kwh REC that I get from that.

So, maybe noise or oscillation is bigger out at the GT array. More microinverters are there, and it's further away and past the A/C load. Maybe I'll try EMI filters at the microinverters, one on each 15a circuit. They could fit in the combiner box.

0 new items by Dale

photos.app.goo.gl

Today's scope video during the issue shows a couple things, I think....
t=0-10s, L1 and L2 look pretty similar, the leads are grounded to the neutral bar, and this is the voltage at the SIs, in the AC1 distribution box. I haven't calibrated the scope voltage, but assuming it's close, it reads about +-160-170 volts, that should be about right. (see on the Ch1 knob, with L1, 5v/div * 10x probe
the frequency is bouncing, the wavelength is changing quickly, same on both legs. the X time axis is 2ms/div, x about 8 divisions per sinewave, so .016 sec period. 1/.016 is 62.5 hz. Again I haven't calibrated the X axis either, but it's about right.
How much is it bouncing? watching the video judging how much the per cycle period is, or freezing it at 9s happens to show it, it looks like about .8ms or +-.4 ms (+- 1 minor division = 2ms/5 = .4s). So, that's +-.4ms / 16ms = +-2.5% and that's 61.5 to 58.5 hz. I could set the microinverters profile further out in freq than this, up 62.5, that might help. Voltages are mostly far out, but the 'slow' limit could be more.
t=20s, lissajou plot doesn't show any phase shifting, the line is fixed slope, it's higher freqs on top of the main 60hz
t=45s, L1+L2 plot shows the voltage difference away from both L1 & L2 being perfectly opposite voltages. The X axis time scale is 2ms/div again. It looks like maybe 12 peaks within the one 60hz cycle (8 div). So whatever is going on is about 720hz? What would that be...
the vertical axis is 1v/div * 10x probe, so it looks like +-10 volts p-p of noise.

I could see what the scope looks like out at the micro combiner box, or the basement panel.

 

[Hedges]

During the afternoon not working time period, the grid-tie micros (connected in beyond the basement panel with the A/C) show all three events, grid instability, ACFreqOOR, and ACVOOR on both L1 L2. The 'off-grid' solar micros connected in at the closer garage panel show only grid instability, ACFreqOOR.

...
the frequency is bouncing, the wavelength is changing quickly, same on both legs. the X time axis is 2ms/div, x about 8 divisions per sinewave, so .016 sec period. 1/.016 is 62.5 hz. Again I haven't calibrated the X axis either, but it's about right.
How much is it bouncing? watching the video judging how much the per cycle period is, or freezing it at 9s happens to show it, it looks like about .8ms or +-.4 ms (+- 1 minor division = 2ms/5 = .4s). So, that's +-.4ms / 16ms = +-2.5% and that's 61.5 to 58.5 hz. I could set the microinverters profile further out in freq than this, up 62.5, that might help. Voltages are mostly far out, but the 'slow' limit could be more.

Not particularly nice "sine wave" of voltage, and jittery.

I tried two Sunny Islands (switched off breaker for other two) with small A/C running and all/some/no Sunny Boys providing PV.

Upper trace is current from Sunny Islands (although scope says volts), and scale was changed for each screenshot (varying portion of current from AC coupled PV), lower is voltage:








The current drawn by A/C is a triangular waveform, but voltage from Sunny Island is fairly sinusoidal. Sunny Boys supply however much current (sinusoidal) and Sunny Island has to make up the balance of the current (which may be out of phase.)

Polarity and selection of one phase current transformer and one phase voltage probe was random, so don't read anything into waveforms being nominally 180 degrees out of phase.

Only first screenshot has excess PV available so frequency above 60 Hz, curtailing production.


For your issue, try shutting off all microinverters, see if voltage waveform cleans up.
Do you have a current transformer to get current waveform?
You may be able to have some PV active without curtailment, but maybe with curtailment they keep connecting/disconnecting causing jumping frequency? What you want of course is smooth frequency shift and power deliver adjusted to exactly match load.

If your inverters disconnect at 59.3 Hz then you should disable Sunny Island's shift to 59 Hz.

 

[ncsolarelectric]

Shucks, still happening. It worked fine with the A/C on for several hours about 11a to 1p, but then later from about 3-5p it's not working.

Today the event logs for the microinverters reveal a couple things of interest.

During the afternoon not working time period, the grid-tie micros (connected in beyond the basement panel with the A/C) show all three events, grid instability, ACFreqOOR, and ACVOOR on both L1 L2. The 'off-grid' solar micros connected in at the closer garage panel show only grid instability, ACFreqOOR.

During the working pretty normal period, both show GridInst and ACFreqOOR, but no ACVOOR. I think these ACFreqOOR events are from the SMAs limiting the solar output, normal behavior with the FLA near full. The grid-tie system shows only one or two ACFreqOOR events; it's freq trip limit is set to 61.3 vs. the default 60.5 of the 'off-grid' micros. The original goal there is to keep the GT system always producing because there's an 11c/kwh REC that I get from that.

So, maybe noise or oscillation is bigger out at the GT array. More microinverters are there, and it's further away and past the A/C load. Maybe I'll try EMI filters at the microinverters, one on each 15a circuit. They could fit in the combiner box.

0 new items by Dale

photos.app.goo.gl

Today's scope video during the issue shows a couple things, I think....
t=0-10s, L1 and L2 look pretty similar, the leads are grounded to the neutral bar, and this is the voltage at the SIs, in the AC1 distribution box. I haven't calibrated the scope voltage, but assuming it's close, it reads about +-160-170 volts, that should be about right. (see on the Ch1 knob, with L1, 5v/div * 10x probe
the frequency is bouncing, the wavelength is changing quickly, same on both legs. the X time axis is 2ms/div, x about 8 divisions per sinewave, so .016 sec period. 1/.016 is 62.5 hz. Again I haven't calibrated the X axis either, but it's about right.
How much is it bouncing? watching the video judging how much the per cycle period is, or freezing it at 9s happens to show it, it looks like about .8ms or +-.4 ms (+- 1 minor division = 2ms/5 = .4s). So, that's +-.4ms / 16ms = +-2.5% and that's 61.5 to 58.5 hz. I could set the microinverters profile further out in freq than this, up 62.5, that might help. Voltages are mostly far out, but the 'slow' limit could be more.
t=20s, lissajou plot doesn't show any phase shifting, the line is fixed slope, it's higher freqs on top of the main 60hz
t=45s, L1+L2 plot shows the voltage difference away from both L1 & L2 being perfectly opposite voltages. The X axis time scale is 2ms/div again. It looks like maybe 12 peaks within the one 60hz cycle (8 div). So whatever is going on is about 720hz? What would that be...
the vertical axis is 1v/div * 10x probe, so it looks like +-10 volts p-p of noise.

I could see what the scope looks like out at the micro combiner box, or the basement panel.

Grid-tied inverters use periodic frequency shifting to detect if the grid is present. If they shift frequency and the grid changes with it, they know it is a resonance and they shut down to prevent exporting power while the grid is down, and electrocuting linemen. If you are trying to AC couple grid-tied inverters to a microgrid powered by an off-grid inverter, the grid-tied inverters are going to continue to do this. I don't know how one would disable anti-islanding protection as it is a requirement for their UL listing, and IEEE 1547 compliance.
If the off-grid inverter has a solid frequency output and is not shifting, then the grid-tied AC coupled microinverters should just think it's the grid. But if the microgrid has a small resistive load and resonances that allow the frequency to shift, they will shut down by design.
You can try adding more resistive loads, like incandescent lights, heaters, and battery chargers. Things that have a unity power factor.

 

[daklein]

Not particularly nice "sine wave" of voltage, and jittery.

I tried two Sunny Islands (switched off breaker for other two) with small A/C running and all/some/no Sunny Boys providing PV.

Thanks for the examples of what they can look like. No I don't have a scope current probe.

Sometimes it looks very perfect and solid.

Here's what it looked like with relatively light loads this morning, and microinverters are providing most of the power and the SI's battery current is less than 5 amps. No A/C running, but there might have been a 120v load like 120v EV charger, or the electric tractor ferro-resonant charger https://photos.app.goo.gl/2B4UBQy9Ms7DosEB8 Not sure if it's L1 or L2 in this one, but the peak gets a little notch taken out of it, intermittently.

EMI filters in various locations could also be trouble for microinverter monitoring. Enphase recommends filters to block and isolate nearby systems.

 

[Hedges]

To see current, all you need is a split-core transformer with resistor, so it outputs voltage.
Mine are like this one, 100A gives 0.333V. Able to see some higher frequencies, a few hundred Hz.

(New) ACCU-CT ACTL-0750-100 Split Core Current Transformer. (100A) | eBay

<p>(New) ACCU-CT ACTL-0750-100 Split Core Current Transformer. (100A). </p><br /><p>Continental Control Systems LLC</p>

www.ebay.com

Sunny Island of course looks like the grid to UL-1741 inverters, which inject some current out of phase with (and/or different frequency from) the sine wave and make sure they can't change line voltage. Not sure if many microinverters working together could disturb the frequency.

I haven't been paying attention to frequency shift with the scope, but DMM Hz scale shows gradually changing frequency. Nothing like your visible jitter, so I wouldn't expect SI to be doing that. Any reason it can't cycle current in/out of battery to make up for staircase power production? Like BMS telling it what current to deliver, so it has to move frequency faster?

It may be you're only having a problem when curtailing excess PV. Get yourself some nice Sunny Boys , or only use UL-1741-SA inverters with frequency-watts enabled.

 

[solar8484]

Thanks for the examples of what they can look like. No I don't have a scope current probe.

Sometimes it looks very perfect and solid.

Here's what it looked like with relatively light loads this morning, and microinverters are providing most of the power and the SI's battery current is less than 5 amps. No A/C running, but there might have been a 120v load like 120v EV charger, or the electric tractor ferro-resonant charger https://photos.app.goo.gl/2B4UBQy9Ms7DosEB8 Not sure if it's L1 or L2 in this one, but the peak gets a little notch taken out of it, intermittently.

EMI filters in various locations could also be trouble for microinverter monitoring. Enphase recommends filters to block and isolate nearby systems.

For the Enphase MI's, the waveform shape is less critical than how jittery the waveform is. It looks like the SI's are struggling with voltage stability control. Can you check the waveform under full/high load *without* the MI's? It would help narrow down the primary cause of the instability.

BTW, given the size and complexity of your system, I would suggest you get better test equipment including the recommended CT. It should be a small investment compared to the cost of the system.

 

[daklein]

Thanks for all the input, I think I'm making progress, it worked better than ever yesterday.

I agree that a current measurement would be interesting, and potentially useful to decide which component is most at fault. It might give clues about a solution, but I can already tell when it is or is not working, by the scope voltage jittery behavior, audible noise, and looking at AC current on the microinverter circuits. Likewise, buying other equipment would be great, but it's a fun hobby to use all this low cost stuff that was almost landfilled or real cheap from DC solar auctions.

It may be related to curtailing, if a large number of micros get curtailed, then they all try to come back on a same time, maybe SI can't handle the disturbance. And a disturbance kicks off any other micros that were not yet curtailed.

Yesterday was a warm day and I ran the A/C again. The last changes I made since last testing:
-added ferrites on all the remaining microinverter circuits.
-a wider limited grid profile in some of the microinverters (picture posted earlier) with the voltage and freq at the limits of the editable range (62.5hz, 141v). For some reason my 2 envoys do not have the same available grid profiles. I suspect they updated the profile template, and only one kept the older template which I had copied and edited. It seems that only that older template available in one of the Envoys will propogate to all the older m-series inverters (either actual old ones or warranty replacements that are calibrated as m-series). I have some other micros which are M or IQ hw & sw, and they seem to accept the 'v2' grid profile or edits of it. I haven't tried to ask Enphase if I can get the other profile added back on the second Envoy. For testing, it's easy to switch the Envoys, scan for new micros and that populates the grid profile if they'll take it. Then mark all those micros as deleted / not active, so it doesn't log their production in the wrong system. Changing back to the other Envoy does then same, but since the v2 profile doesn't take, they are left with the other profile. Kludgy but seems to work.

Anyway, it worked all day; With the 62.5hz profile, in the GT and battery micros, with the A/C running & rest of the house on the SunnyIslands. For most of the day, I left the OG micros and the garage load panel coupled to the grid, and let that excess go out.

Later in the afternoon I tried running the garage load panel with the OG micros also off-grid on the SunnyIslands. For a while it worked, but the OG micros did not all produce the nominal amount they should have. The frequency control issue started happening again, and all the microinverters stopped. This could potentially have been triggered by curtailment. It was later in day, house was already too cold, and so I didnt' confirm if taking away the OG micros with their default 60.5hz profile would fix it.

So next step, for next hot day, I put the 62.5hz profile in the OG microinverters. So all should have that wider profile, except three M series inverters on the OG garage array. They would revert to the v2 template profile, 61.3hz limit, that I last tried to put in the OG micros. This could be helpful for curtailment, at least those 3 will go off first. I also removed the ferrites from the OG garden circuit, because it seemed like the Envoy was not finding those micros, but it may have just been evening and micros already sleeping for night. It found them and updated the profile fine this morning.

 

[Hedges]

Wide frequency range may help so long as microinverters ramp down power in response to frequency shift.
If they simply disconnect, all that happens is frequency shifts further and they disconnect a second or two later.

I asked someone before if each microinverter could be programmed to disconnect at a different frequency, but apparently Envoy propagates settings through to all, can't be different. Also, need instant reconnect, not after a 5 minute wait, while frequency ramps up and down within a certain range.

That jitter - seems likely to be the issue, whatever causes it.

Do those microinverters have frequency-watts function? Do they have volts-watts function? I haven't heard, but wonder how volts-watts would perform.

If problem is related to frequency swinging too far before they respond, possibly if another well-behaved inverter like Sunny Boy was also connected, it would absorb variations in frequency/power by responding smoothly. That could slow the shift enough to let your dumpster-diving collection of microinverters function.

 

[daklein]

Exactly, it's a dumpster diving collection! haha! And you'd only do this as a hobby.

No, the m or M series do not have freq-watt or volt-watt feature and there's no grid profile to use that would add it. I asked Enphase a while back. I don't know of any way to have them ramp down. (little m I'm calling the oldest aluminum case m190/m210 and their warranty replacements with newer M or IQ hardware, big M is the aluminum plate on one side, black alum housing on other side)

Right, I don't know of a way to set different grid profile for micros within one system. But I have two envoys and systems, so that's two sections, and the OG system can have two within it, by using a grid profile that only some micros will accept because they are different h/w or s/w (the M215s I have on the garage roof). Once there' s some excess (freq getting to where I had the OG system set), the GT system is hooked back to the grid automatically. And there are various loads in the house which are controlled depending on the excess.

Right, wider freq setting won't change curtailment. When it gets there, they all turn off. But, if it gets there sloowly enough, it works fine because not every microinverter trips off together. There's some variation in measurement of the frequency by all the micros, some trip first typically. Except for running the A/C, it all works well and no issues when it's curtailing. The monitoring shows that microinverters turn off somewhat randomly during curtailment (it's only 15min interval production data).

I think the issue is the fast-changing freq, which looks like jittery voltage and makes a light growling from the SIs. Once it's doing that, the micros will not connect to it, it looks like it is +-3-5hz from 60. If the wider tolerance grid profile can avoid micros getting tripped by whatever initial disturbance or instability of the system or the SIs control, it may work. It seemed to work yesterday for the micros with the 62.5hz profile. However, I also did have the EV charger at 2.8kw for some of the time with the AC, so maybe that made the system have a better PF or affected the interaction of the micros and the SIs.

@ncsolarelectric described the anti-islanding feature of grid-tie inverters in general. Does anyone know if this is done continuously, or if is only once the inverter connects again? If that diagnostic is being by a bunch of micros at same time as they are trying to restart, with the right system loads and sensitivities, maybe that sets it off and the SIs control continues it.

On the other hand, it could be a temperature issue with the A/C load, or the SIs, or the microinverters too.

So I'll see the grid profile change to the remaining OG micros improves it, next time it's hot. Any other theories or experiments to do are welcome, especially easy ones.

Do the Sunny Islands have any hidden parameters or calibrations that might have an effect? Is there a list of all the direct access parameter register numbers with what they do, I'm not sure they are all listed in the manual.

 

[Hedges]

Anti-islanding is continuous. The issue would be if grid was up, and later a wire broke so you just powered your house and the neighbors, so line remains hot.

Can't microinverters operate without Envoy? What if you disconnected AC (or possibly just DC?) wires from all but one, set it for 60.1 Hz. Disconnect it, connect one other, set for 60.2 Hz. Etc. Then disconnect Envoy and reconnect all microinverters so all have different settings.

Sunny Island manual appears to include all parameters.
The Sunny TriPower I just got, manual doesn't list any parameters.
It appears to me that when I use Sunny WebBox, all I need is administrator password for WebBox. Don't need inverter password, installer password, grid-guard code, anything to set any and all parameters! (at least for Sunny Boys.)
With Sunny Boy Control, password was sum of digits of date. With Sunny Explorer and Speedwire, password appeared to be stored in the Speedwire module, and Sunny Explorer enforced updated password complexity rules.

One key parameter to set in Sunny Island prevents it from dropping to 59 Hz for the purpose of getting mechanical clocks back on time. That would trip UL-1741 inverters with 59.3 Hz limit.

Any load that draws more power at higher frequency should help. Motors, like pumps, do. Of course your AC does. But it would be nominally 1.5%/Hz, vs. 100%/Hz with frequency/watts curtailment. SMA once had Sunny Load or something like that, a frequency-shift controlled dump load.

 

[daklein]

Can't microinverters operate without Envoy? What if you disconnected AC (or possibly just DC?) wires from all but one, set it for 60.1 Hz. Disconnect it, connect one other, set for 60.2 Hz. Etc. Then disconnect Envoy and reconnect all microinverters so all have different settings.

Yes, that could work actually, good idea. I could do that for the battery micros. As long as you don't want to monitor power production with it, and I do because otherwise it's not obvious if they fail. You can scan for micros first, with all connected, and it will get them in its list and they'll be 'active'. The grid profile only goes to micros that are selected as active in the Envoy. Check a few micros as 'deleted' (they're still listed in the Envoy, but don't get monitored or sent the grid profile), change the grid profile, let those update, delete a couple more, change profile again, etc.

Sunny Island manual appears to include all parameters.

For example the instructions about how to do grid export from the SIs, says go to direct input 600 menu to 27101 setting to enable battery feed in. There are parameters for export target reactive current (in normal menus, once the export is enabled with 27101). I wonder if there are any other settings in the 600 menu, that might change behavior like power factor correction, which might affect this voltage and frequency control.

One key parameter to set in Sunny Island prevents it from dropping to 59 Hz for the purpose of getting mechanical clocks back on time. That would trip UL-1741 inverters with 59.3 Hz limit.

Yes, that's turned off. Very funny to have that luxury feature, but maybe not robust in this other way. But the SIs are neat for sure.

My system controls, the RPi, looks at the SI reported frequency, main FLA battery current & voltage vs. target voltage, to turn on more loads when there's excess, attempting to put the load on before the micros get tripped off. I don't have anything that directly measures the frequency. I'd like to make my 240v EVSE do variable charging which would have plenty of control authority and could be finely adjusted. I haven't gotten to that yet, maybe that could measure the frequency directly.

 

[solar8484]

Exactly, it's a dumpster diving collection! haha! And you'd only do this as a hobby.

I think the issue is the fast-changing freq ...

This is the reason why I mentioned the jittery voltage waveform is the more serious problem than the waveform shape ... Enphase anti-islanding detection includes continuous checking of RoFC (rate of frequency change) and it does not appear to be affected by the frequency range settings ... even worse Enphase checks not just measurements at the fundamental frequency (i.e. 60Hz) but also the harmonics ... I suspect this is a key reason why people appear to have more AC coupling problems with Enphase compared to other GTI (e.g. Sunny Boy).

@ncsolarelectric described the anti-islanding feature of grid-tie inverters in general. Does anyone know if this is done continuously, or if is only once the inverter connects again? If that diagnostic is being by a bunch of micros at same time as they are trying to restart, with the right system loads and sensitivities, maybe that sets it off and the SIs control continues it.

Enphase anti-islanding detection is done continuously and it's done via multiple mechanisms including active ones such as frequency shift and harmonic current injection. It's essential that your SI's are able to maintain island grid voltage stability with the disturbances. Your scope traces show otherwise. So, as previously suggested, it would be good if you can first verify your SI's can maintain stable island grid voltage under load but without the micros to confirm you have a solid foundation.

On the other hand, it could be a temperature issue with the A/C load, or the SIs, or the microinverters too.

So I'll see the grid profile change to the remaining OG micros improves it, next time it's hot. Any other theories or experiments to do are welcome, especially easy ones.

Thermal damage is certainly possible. I've seen reports of Enphase micros (mainly iQ series) that have suspected thermal damage causing regular drop outs. You can check by removing the micros that have the most frequent drop outs to see if that helps.

 

[daklein]

Do the Sunny Islands have any hidden parameters or calibrations that might have an effect? Is there a list of all the direct access parameter register numbers with what they do, I'm not sure they are all listed in the manual.

The dumpster flies got stuck in my bonnet so I looked in the direct access menu 600, 27xxx parameters. There are a few! Is anyone familiar with the SMA abbreviations, or know of any documentation?

Maybe these could be relevant.
27104 Frq Drp -0.2 Hz/kW
27105 Vtg Drp -0.68 V/kVAR
27129 Inv Drp Ctrl Normal
27112 Lod Frq Ena Off
27113 Lod Frq Del 3 Hz
27114 Lod Frq Rmp 0.1 Hz/sec


Here's the whole list. I didn't see anything else after 27129
27101 Gr Fed Ena disable
27102 Si Com Ena disable
27103 Si Com CAN intern
27104 Frq Drp -0.2 Hz/kW
27105 Vtg Drp -0.68 V/kVAR
27106 Div Frq Min 61 Hz
27107 Div Frq Max 62 Hz
27108 Div Pwr Max 0 kW
27109 BMS Mode Auto
27110 Gn Rt Cur Src External
27111 Ext Cur Sns 90 A/5A
27112 Lod Frq Ena Off
27113 Lod Frq Del 3 Hz
27114 Lod Frq Rmp 0.1 Hz/sec
27115 Ext Cur Rt 0 A
27116 Long Ch Lst Ena Disable
27117 Bat Di Chg Cur Ma 1200 A
27118 Bat Di Chg Vtg Mi 35 V
27119 Sic Bat Vtg Del M 0 V
27120 Bat Ety Ofs 0 mV
27121 Num 4
27122 Mc Box Con Gen
27123 Gn Tm Ena Disable
27124 Gd Tm Str 00:00:00
27125 Gd Tm Stp 00:00:00
27126 Grd Apt Phs Disable
27127 Ext Cur Rmp Slp H 18200 digit/sec
27128 Ext Cur Rmp Slp L 1820 digit/sec
27129 Inv Drp Ctrl Normal
27130 not found
27131 not found
27132 not found
27140 not found
27150 not found
27200 not found

 

[ncsolarelectric]

Exactly, it's a dumpster diving collection! haha! And you'd only do this as a hobby.
....

@ncsolarelectric described the anti-islanding feature of grid-tie inverters in general. Does anyone know if this is done continuously, or if is only once the inverter connects again? If that diagnostic is being by a bunch of micros at same time as they are trying to restart, with the right system loads and sensitivities, maybe that sets it off and the SIs control continues it.

On the other hand, it could be a temperature issue with the A/C load, or the SIs, or the microinverters too.

So I'll see the grid profile change to the remaining OG micros improves it, next time it's hot. Any other theories or experiments to do are welcome, especially easy ones.

Do the Sunny Islands have any hidden parameters or calibrations that might have an effect? Is there a list of all the direct access parameter register numbers with what they do, I'm not sure they are all listed in the manual.

It's repetitious, but not continuous in string inverters. I don't know if the Enphase micros work in sync or not. I know 3 phase string inverters are not synchronized unless you create a network, which isn't a requirement for anti-islanding to work, or for them to behave properly. I would think in general, that the micros are not in sync, but if they are I can't imagine why. The whole objective is to detect that the grid is "stiff", not to intentionally make it less stiff which is what I think would happen if they all synced up.

 

[Hedges]

The dumpster flies got stuck in my bonnet so I looked in the direct access menu 600, 27xxx parameters. There are a few! Is anyone familiar with the SMA abbreviations, or know of any documentation?

Maybe these could be relevant.
27104 Frq Drp -0.2 Hz/kW
27105 Vtg Drp -0.68 V/kVAR
27129 Inv Drp Ctrl Normal
27112 Lod Frq Ena Off
27113 Lod Frq Del 3 Hz
27114 Lod Frq Rmp 0.1 Hz/sec

The manual of course: https://files.sma.de/downloads/SI4548-6048-US-BE-en-21W.pdf

600# direct access is addressed, but I don't see 27xxx

"27104 Frq Drp -0.2 Hz/kW"

I thought frequency vs. kW was a percentage for Sunny Boys. Off-grid setup 1.0 Hz swing from 61 Hz to 62 Hz drops power output from 100% (of available) to 0%. I think that is of available at the moment, not of rated inverter power, but not certain.

This parameter value appears to be 1.0 swing (100%) for 5kW?

Found it in 5048US manual, page 138! This manual "Installation and Instruction" is different from another I stored previously, "Technical Description"

https://www.altestore.com/static/datafiles/Others/SI5048U_Manual.pdf

"271 01 FrqDrp Hz/kW -0.2 frequency droop (PF)"

However, Others I don't find such as:

"27107 Div Frq Max 62 Hz"

 

[daklein]

"27104 Frq Drp -0.2 Hz/kW"

I thought frequency vs. kW was a percentage for Sunny Boys. Off-grid setup 1.0 Hz swing from 61 Hz to 62 Hz drops power output from 100% (of available) to 0%. I think that is of available at the moment, not of rated inverter power, but not certain.

This parameter value appears to be 1.0 swing (100%) for 5kW?

Found it in 5048US manual, page 138! This manual "Installation and Instruction" is different from another I stored previously, "Technical Description"

Thanks! some of the manual versions are little different, and the s/w would be different over time too.

It seems like the freq-watt or SMA's name FSPC Freq-shift Power Control is configurable in the Boy and the Island. The SMA manuals describe how to set the SB. And the default settings in my SI, at least for 27106,107 of 61 & 62Hz correspond to the SI operating manual figure on sec 17.5 p 155.

The 5048 manual gives a longer name to 27104 FrqDrp Frequency Droop (PF) and 27105 VtgDrp Voltage Droop (QV)
I wonder if these two are for grid feed outflow operation, GrFedEna=1, to specify how to change the SI outflow power if & when grid or cluster freq & voltage goes away from nominal. Analogous to how a solar PV inverter does freq-watt.

27129 InvDrpCtrl settings available are Normal, FrqCtrlOff I imagine that is probably the FSPC freq-watt feature, or maybe it's the droop control when connected to AC2.

This sort of topic: https://www.sciencedirect.com/topics/engineering/droop-characteristic

I haven't run the A/C again yet for that issue, but I notice a better behavior with my change to the grid profile on the rest of the microinverters. Previously, a bunch of the micros would trip at their default 60.5 hz. The SIs seem to be set up to ramp between 61-62 Hz (I might want to change that upper to 63Hz now, but it may not matter, I have seen them go up to 64 or 65Hz when the DC coupled solar target voltage was not matched with the SIs....) What seems to happen is that the SIs ramp from 60.0 up very quickly, and don't typically keep the FLA battery charging very much even, and that would take out many of the microinverters. Sometimes the SIs will jump straight up beyond 60.5 and it's too late. Between 60 and 61 is only no reduction required and just more than zero. Now with all the micros set higher except three at 60.7 slow, 61.3 fast, it ramped up and those went off, with freq stopping at 60.7

 

[Hedges]

Sunny Island doesn't distinguish 61 or 62 Hz. It knows 60.5 Hz from UL-1741. Off-grid it ramps up to 64.5 or 65 Hz if necessary to curtail production (settles between 61 and 62 Hz when Sunny Boy responds properly). I saw that when Sunny Boy 10000TL-US set for grid-backup failed to ramp down power output. It only implemented higher frequency limit 64.5 Hz and I think reconnected without 5 minute wait. (Off-grid settings made it work as desired, and 5000US I know use does grid-backup correctly.)

 

[daklein]

I tested today. With all microinverters set to 62.5Hz, 10s reconnect time and plenty of sun, the system settled at 62.4Hz according to SI display, while the SI battery voltage was near the target float voltage and charging appropriately gently.

 

[Hedges]

While that sounds good, is that just a corner case where either some/all microinverters disconnected, or slightly higher frequency increased power draw of loads?

Need to try a wide range of loads, of available PV production, of battery SoC.

If you can set each microinverter to a different max frequency, maybe shorten reconnect time, it could more smoothly ramp up/down.

For SolArk, it appears having GT PV connected behind the "generator" relay, and a greater amount of DC coupled PV, it is supposed to be well behaved. It can gradually ramp down inverting from DC to AC, then kick PV offline and replace lost power with DC, ramp down again. Maybe you can get similar behavior with enough DC coupled PV.

 

[daklein]

Update: SMA service has responded to my service ticket and they seem to be thinking about it. No good advice yet, but better than more than a month of silence. I made some better video clips of the issue occuring, if anyone's interested. https://photos.app.goo.gl/dYe4DZsSzcGr4B1VA The info tab on each picture or video has some description.

This patent, https://patents.google.com/patent/US7899632 might help explain the issue. I wonder if the SMA inverters are not always robust to this anti-islanding behavior from the micros. Just under certain conditions of loads, higher system load from the A/C, car charger, etc, but with the SI inverters not really doing much at all, since all the microinverters are supplying the loads.

With a couple of changes, it works better sometimes. Sometimes perfectly for hours with the A/C running, but other times still oscillates intermittently and stops AC coupling. If it's working fine with the A/C running, with all the micros happily AC coupling, if I add or remove other loads like the 240v EV charger or 120v mower battery chargers, the change in load can make the issue start or stop happening.

Two changes that seem to matter, without both of these, the issue usually occurs right way:
The microinverters have a wider grid profile (62.5Hz on all of them, except 3 are at 61.3Hz, long story). The wider tolerance (vs. the 60.5hz default microinverter grid profile) allows a little more room to not trip them all off.

I set 27104 and 27105 to values at or near zero instead of the default values. Have to _not_ try to connect to the grid (for charging or grid-feed), with these settings. It will initially synchronize with the grid, but within a couple seconds it opens the relay and gives a fault for AC2 shorted.