diy solar

diy solar

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

daklein

New Member
Joined
May 8, 2020
Messages
190
I have a system mostly as shown in the attached diagram. Recently have added a separate ~13kwh 14s lithium battery bank which the Midnite CC charges, and 10 microinverters enabled by a relay board discharge the lithium battery mostly at night, or any time there are more loads like the AC load. For a year & half, I've been using 4 microinverters to discharge a ~9kwh 36v lithium battery bank, just at night, the same idea, to avoid as much cycling of the FLA battery. It generally works great although there's always something to adjust in the RPi that controls it all; the FLA battery generally stays full floating, except when larger loads turn on, until the right number of microinverters are turned on to match it.

The conventional old air conditioner is located right outside the basement panel, and the basement to the garage inverters run is maybe 60 feet of #6 copper. The A/C seems to take 4-5kw. I tried increasing the nominal AC voltage of the SIs to 122 from 120, and maybe

The enphase inverters have been set up with different off-grid profiles, one set with a little wider frequency tolerance, so not all of them drop out at once when the SIs ramp up Hz to manage the FLA near full.

When I run the air conditioner, sometimes, the microinverters will not always stay producing. Last summer I did notice a sort of light, variable growling or rumbling noise from the SIs when the AC was on, but everything still seemed to work well. At that point, the 9x305w panels were DC coupled to the FLA as shown in the diagram. I'm wondering if that helped somehow, providing extra DC power to the SMAs, at least in the day. Now I am seeing sometimes that the microinverters are not producing, both day (plenty of AC coupled solar available), and night (a bunch of microinverters discharging from lithium batteries). The nominal reported Hz and voltage are fine, but the data plots are pretty slow rate data every few seconds at the most, so the plot data may be somewhat random-looking aliasing of whatever faster behavior is happening.

In the plotted data from the evening, there are several periods of the A/C running: 20:20 - 20:35, 20:50-21:00, 21:35-21:40, and 22:20-22:25 which I'll call periods 1,2,3,4.
The first period, 20:20-20:35, and 4th period 22:20-22:25 seem to not have the microinverters operating very well. The reported freq is 60hz, +- .05 (green line 4th plot down), and the micros should all stay on until 60.5 by default or 61.3 that I have some set to. Actually, now I see that the 'min' & 'max' values are a bit wider. These are from pretty fast about 10Hz I think CAN data and keeping the min & max over last 75 values. The lithium battery voltage that the micros are taking from (blue line top plot) is noisy and not so much decreased from it's value a minute before when the micros were all off.

Maybe I should be checking the A/C compressor start & run capacitors? I'm not sure why it's intermittent.

I'm wondering about adding a 1kVA 240:32v transformer, it weighs 15-20 pounds, to help smooth the voltage at the basement breaker panel, or maybe that should be close to the microinverters? Can I add some sort of filter to help the micros see a cleaner waveform? Should it be located at the A/C compressor, or should it be located nearest to the microinverters?

One day we'll get rid of the old PSC compressor motor A/C and get a nice variable speed inverter driven geo or airsource heatpump, that'll be easier to run cleanly. Any other lower cost ideas or suggestions please? Thanks!
 

Attachments

  • GT_OG_Battery_solar_system_wTrnsfrSwtch.jpg
    GT_OG_Battery_solar_system_wTrnsfrSwtch.jpg
    118 KB · Views: 22
  • Screenshot_20220617_001426.png
    Screenshot_20220617_001426.png
    213.8 KB · Views: 20
Today I checked the run capacitor in the outdoor AC compressor / condenser unit. It checked fine, but it could be 30 years old so I swapped in a different one. No change, the SunnyIslands still go between a mild grumbly growling sort of sound or normal hardly any sound, and the microinverters not producing when it's grumbling. It goes in about 5-10 minute cycles, starting at 4:52 when the AC was turned on, until about 5:30 it stays grumbling until 6:15 when the AC turned off. During the grumbly time, the SI battery current is high, and more battery microinverters turn on (the solar ones are not producing and neither do the battery micros). Then suddenly for some reason the micros all work for a few minutes, with the battery microinverters turning back off as the main battery is charging.

The behavior starting at 6:40 is normal, this is the induction stovetop and oven cooking dinner. The RPi sees more load (SIs discharging FLA battery) and turn on more lithium battery microinverters (orange line top plot) to get the main SI current back close to zero.

I read about buck/boost transformers, which I have one of laying around. https://www.carltonbates.com/static/catalog/products/images/PDF/T137921.pdf (not this brand, but this is the idea)
Maybe putting one of these on the AC, wired to boost it by 16 volts, 240:256, would help out the compressor motor with a little higher voltage and maybe keep the waveform from getting too out of shape?

And another simple thing to try, add some ferrite cores, can't hurt much? I have a bunch of cores from old Enphase inverters, and I put three on each L1, L2 AC1 from the SIs, and two on the L1L2 circuit going to the battery microinverters. The picture is the AC1 junction box from the Sunny Islands, with the 50a breakers each going to a house panel, one of which has the AC on it.
 

Attachments

  • Screenshot_20220620_194755.png
    Screenshot_20220620_194755.png
    675.3 KB · Views: 15
  • IMG_20220620_224353322.jpg
    IMG_20220620_224353322.jpg
    377.9 KB · Views: 15
In your diagram, "A/C blower" appears to be on grid only, not powered by the Sunny Islands and Enphase. But you say it interferes.

I've used a scope to capture current waveforms for various loads powered by SI.
A/C I expected a phase-shifted sine wave, but it was a triangle.
VFD running pool pump is what upset my GT PV inverters. I didn't capture voltage at same time (I was studying battery current) but I can imagine the distortion.


If your microinverters share a wire run with A/C, current waveform would produce distorted voltage waveform.

I don't think the ferrites will do much at low frequency.
The ones around a pair of wires will pass differential and not saturate if equal current. That leaves their inductance available to block common-mode, especially high frequencies.
The ones around a single wire ought to be saturated and ineffective. Although, it has been reported to fix an EMI issue for a compliance test.

Maybe a transformer feeding compressor will reshape the waveform a bit, not passing high frequencies as well and starving it for current if it wants peaks.

Do you have oscilloscope images?


"... get a nice variable speed inverter driven geo or airsource heatpump"
That'll only be OK if it has a high power factor. If a simple rectifier/capacitor front end like mine it will be worse.
I've bought PF correction boost converters but not tried them yet; need some support circuitry.
 
In your diagram, "A/C blower" appears to be on grid only, not powered by the Sunny Islands and Enphase. But you say it interferes.
whoops, that's history from when I left the AC on the grid, and the compressor was on another separate grid meter. Now everything is on the off-grid panels, blower & compressor.

I've used a scope to capture current waveforms for various loads powered by SI.
If your microinverters share a wire run with A/C, current waveform would produce distorted voltage waveform.
Yes, I should figure out how to run the nice scope I've inherited... And I should draw a diagram showing the potentially relevant wire runs through-out the system.
The SIs are in the garage, with about 10' run to the now 18 battery microinverters in garage and 3 solar micros on the garage roof. About 150' over #10 wire are 13 solar micros in the garden. About maybe 60' to the basement panel with #6 wire with the AC compressor outside 6' away, and the grid-tie roof array is about 30' #6 to the transfer switch and maybe 50' of #10 up to the 21 solar micros.

Thanks for your suggestions
 
Multiple groups of microinverters on different branches? Any statistics on disconnect issues vs. which branch? Vs. which model?
Any microinverters sharing the 60' #6 run to AC compressor?

Presumably voltage waveform, voltage/frequency/waveform (harmonics) are what affect the inverters. Or impedance, since UL-1741 involves driving out of phase/frequency pulses to confirm a bottomless grid is connected (which you might see as voltage on the scope if not low impedance.) Current transformer (with resistor so converted to voltage) is interested to see current waveforms.

I had switched from 10000TLUS to 5000US thinking transformer would isolate it from the noise, but still had upsets. For now I avoid operating VFD off-grid, but it can happen if grid goes down. Future will wire separately or get PF correction circuit going.

Maybe a "run" capacitor would improve waveform from AC compressor. But probably more at some level of load than another.

You could also have voltage excursions as compressor starts & stops. Check what voltage you've got. Buck/boost for inverters could shift that.
 
Have you ever thought of a hybrid inverter to control it all?

You close to Ann Arbor? I got a real oscilloscope if that would help you.

I don't see in the Enphase aps where to change the micro inverter settings with the new envoys. I see you have the old Envoy, what program do you use? I like the idea of changing the frequency kick out points.

Thanks
 
There are certainly transients when the AC starts up, enough to cause trouble with items on the 120:24v control transformer. It would sometimes start the AC fine, but other times reset the thermostat or potentially not hold the AC contactor on.

I have a separate old APC UPS with LiFePO battery that runs the internet & network devices, which I will add the control transformer on at some point. Until then, during cooling season, I have the control transformer plugged into the grid. This way the AC starts fine, just a little bump sound from the SIs, but I'm sure it's totally within their capability.

Here are some scope pictures. I think the scope solved the issue because today it's working properly so far. No grumbly noise at the SIs and all the micros seem to be working. The waveform doesn't look too crazy? I can't see any difference with and without the AC on so far, but I'll try to catch it 'complaining'.

@newbostonconst On the Envoy, under grid configuration, then manage profiles, the profile can be edited. There are different available profiles in the two different older Envoys I have. I think you may be able to contact enphase and they can load on different or updated grid profiles to the Envoy. Both allow me to edit the profiles on the local admin web page, but only my original installation envoy seems to be successful in updating to the micros. The other one I got used and was able to get it set up as a new site, They both report having same s/w 3.7.31. The used one is for the 'off-grid' array and I leave that at the default config, 60.5hz trip threshold. It also required logging into an installer Enlighten account to get a code to update the profile on the Envoy.

I wanted to keep the SIs and the FLA as the main system, but not cycling the FLA too much. The microinverters were all basically free. But yes, it would be neat to use some other smaller inverter to AC couple the lithium batteries, maybe one with a GT limiter type of feature. Do you know of a system to combine multiple different battery banks.
 

Attachments

  • Screenshot_20220621_125654.png
    Screenshot_20220621_125654.png
    123.3 KB · Views: 5
Have you researched combining the two types into one bank.

Lead acid has to go below 12.6 volts before it really is starting to discharge.
LFP is 12.0 completely discharged 12.8 nominal and 13.5 charged.

Do you have two shunts? You could put one shunt on the lead acid and one shunt on LFP and watch what they do when hooked together. I bet LFP would do almost all the work.

Your Sine wave looks much better then my old outback.....

I have a MidNite classic on a 4x4 array 16 panels and switched out the 250 for the 200 and gained 20% daily output. The 250 isn't as efficient, I got suckered into the thought process of it cost more it must be better.....nope.

With the new envoys I can't find where to change there settings other then the generic grid profiles. Thanks
 
Have you researched combining the two types into one bank.

Lead acid has to go below 12.6 volts before it really is starting to discharge.
LFP is 12.0 completely discharged 12.8 nominal and 13.5 charged.

Do you have two shunts? You could put one shunt on the lead acid and one shunt on LFP and watch what they do when hooked together. I bet LFP would do almost all the work.

Your Sine wave looks much better then my old outback.....

I have a MidNite classic on a 4x4 array 16 panels and switched out the 250 for the 200 and gained 20% daily output. The 250 isn't as efficient, I got suckered into the thought process of it cost more it must be better.....nope.

With the new envoys I can't find where to change there settings other then the generic grid profiles. Thanks
They are NMC Leaf cells, not as potentially compatible to combine in one big bank as LPF w/ FLA. And the SIs would have no idea about the FLA SOC anymore. There is one shunt on the FLA that the SIs use. No shunt on the lithium packs, beyond the BMS current and voltage data on each pack. I use only pack voltage to control charge and discharge, inside of what the BMS is set for.

I found a loose connection that might matter. It was loose enough to pull out of the breaker, on L1 L2 from the AC1 distribution box going to the basement panel with the air conditioner.
 
Ok, it still does it. I added a couple more scope pictures to the goog photos album linked above, and here's a video of it happening. https://photos.app.goo.gl/EVKaqwMbYXP1JZEp6

It seems to be when there is solidly extra AC coupled solar available than what the house and AC load requires. Maybe the times it doesn't occur is when there is not so much excess.

The sine wave looks normal when plotting both L1 L2 separately, but the wavelength or frequency does jump around some. With the L1 + L2 summed (the voltages should normally sum to zero, no?) the variation is more obvious.
 
There are certainly transients when the AC starts up, enough to cause trouble with items on the 120:24v control transformer. It would sometimes start the AC fine, but other times reset the thermostat or potentially not hold the AC contactor on.

Sounds to me like the A/C is too much load, if line voltage drops sufficiently that thermostat or relay is affected.
(expensive) soft-start kits are said to reduce current surge greatly.
Measure the voltage droop and starting current.

I see the sine-wave traces, don't know what the low amplitude ripple noise and triangles mean.

Can you trigger on starting current, with current probe or current transformer?
 
Sounds to me like the A/C is too much load, if line voltage drops sufficiently that thermostat or relay is affected.
(expensive) soft-start kits are said to reduce current surge greatly.
Measure the voltage droop and starting current.

Can you trigger on starting current, with current probe or current transformer?
The A/C starts fine since last year figuring out that powering the control transformer separately is good. It seems like it starts in half a second, by the burupp sound at the inverters. It would be good to have a start kit on the A/C, but the general plan is that this A/C is temporary. The compressor motor plate says 95 locked rotor amps x 240 = 22.8kw which is pretty close to the 2x11kw 3 second limit.

It takes 3-4kw when running, and AC coupled solar is 2 times that when there's sun. Even if the startup disturbance causes microinverters to drop out for a short time, the trouble is with keeping the microinverters happily AC coupling all the time. Seems like 2 SI6048 should make childs play of this A/C unit.

I'm looking for ideas about what could make the SI frequency jump around, even the nominal frequency is still 60.0hz. This seems to be preventing microinverters from working.

Maybe a power factor correcting capacitor at the A/C compressor would help. According to the SI rs485 data, with the A/C running at night (no solar microinverters to help, and some of the battery micros turning on intermittently), the total apparent power is about 4kw, and the reactive power is 400-800 watts. So is that power factor of 4000-800 / 4000 = .8 PF ?

Here's the A/C unit diagram. It has only the dual run capacitor for the fan & compressor, with start thermistor across the compressor cap as shown. It doesn't have the start relay and start capacitor.

Maybe I need to quit with this old A/C unit and get something modern. I'll have to be careful about that too that the SIs will like it.
 

Attachments

  • IMG_20220620_204950278.jpg
    IMG_20220620_204950278.jpg
    425.9 KB · Views: 0
So per spec it should work.

Power Factor is real power / apparent power. That's one leg and hypotenuse of Pythagorean theorem, and reactive power is 90 degrees to real power:


Microinverter won't know above power factor of load, just the voltage waveform. So frequency, voltage, harmonics, stability. I would expect A/C to run steady-state, not change wave shape rapidly most of the time.

SI does make excursions down to 59 Hz periodically, by default. You can tell it not to.
Below 59.3 Hz knocks old UL-1741 off line. With UL-1741-SA should stay on line full power between 58.5 Hz and 60.5 Hz, ramp down above that if frequency-watts supported.
Frequency excursions also occur when reconnecting to grid.
 
Here's an updated system diagram for entertainment, or ideas. I opened a ticket with SMA, see if they have ideas. I did also add a 521 soft start to the A/C unit, it should be better than the start thermistor shorting the run capacitor. I wondered if the start thermistor could somehow fail to low resistance intermittently and remove the run cap while running, seems not. The A/C seems to start about the same, just a short burrp from the inverters.

I wonder about the lengths of the wire runs, and where the micros & loads are located, I don't know if the topology matters. For example, the A/C load is ~80' away, and the microinverters are beyond that maybe 120'. Does anything happen when the maybe normal A/C running load of a couple kw is suddenly negated by the microinverters.

Experimenting again today while running the A/C, I found that the noise came and went if I turned off the battery microinverters. But I didn't see all the solar micros start to produce and bring down the SIs FLA battery discharging. I wonder if it's not really the A/C load, but just the total amount of load on the SIs, and some interaction with a bunch of microinverters all trying to hop back on at once.

An undesireable option is to put both load panels on the grid manually. This puts the A/C and all the AC coupled solar on the grid and if there's more than hazy sun it's enough to run with excess going to the grid. But if another load like the kitchen stove comes on also, it pulls from grid, when there's plenty of battery and more solar in the remaining off-grid side. And I do enough running around already.
 

Attachments

  • system_diagram_0622.jpg
    system_diagram_0622.jpg
    133.5 KB · Views: 10
Last edited:
Possible experiments:

Measure voltage at microinverters. (your meter may measure voltage different for anything other than pure 60 Hz sine wave.)

Replace A/C load with similar wattage resistance heaters. That would be clean sine wave load.

Feed microinverters through EMI filter.

Transformer isolate microinverters.

Buck/boost change voltage seen by microinverters.


Depending on wattage/current rating of the parts available to connect in series, you could connect fewer microinverters .
EMI filter mostly blocks higher frequency common mode.
Transformer should attenuate somewhat higher frequency differential mode, because core doesn't transfer it well and will have lossy eddy currents. Wired for isolation it would block common mode.
Are the microinverters 120V? Any transformer configurable for 120/240V primary could be used as 120V isolation transformer.

Voltage tolerances should be wide enough that some droop due to current feeding A/C (once running) shouldn't be enough to knock inverters offline. Especially since Sunny Island delivers solid 120V. Utility line on the other hand can be set to run high, and inverters push it close to high voltage limit.

I would expect poor voltage waveform, higher frequency harmonics, are what makes microinverters disconnect. But I'm not sure how they process waveform; UL-1741 specs just say voltage and 60 Hz +/- frequency. Maybe distorted waveform gets read as different RMS voltage. Is there a status reporting what disconnects them or keeps them offline? My string inverters report voltage or frequency which causes that. Any transformer could be used as buck/boost to shift what they see e.g. 120V + 6V = 126V, 120V - 6V = 114V. If the RMS voltage it sees is out of spec, one of those to adjustments could bring it back into range.
 
Have you looked the Envoy event logs when the microinverters stop producing?
I had not looked at the logs recently until now, good idea. I ran everything off grid again today (it worked better today somehow...)

When it's ringing / oscillating / growling, it will set AC Frequency Out Of Range and Grid Instability. However, it will also say these normally, whenever the SIs raise the frequency to reduce AC couple power input.

There are also events today for AC Voltage Out Of Range - Phase 1 and 2, but I'm pretty sure these are when I was switching the load panels from off-grid inverters to grid or vice versa. The grid-tie system with it's automatic transfer switch also will set these.
 
Possible experiments:

Measure voltage at microinverters. (your meter may measure voltage different for anything other than pure 60 Hz sine wave.)

Replace A/C load with similar wattage resistance heaters. That would be clean sine wave load.

Feed microinverters through EMI filter.

Transformer isolate microinverters.

Buck/boost change voltage seen by microinverters.
Today I ran the whole house again off grid, with the A/C, the oven cooking off & on, and charging the EV some. It did still growl for a few short periods, maybe 1 second at a time, but it did not continue, and the microinverters seemed to spend most of the time working. I'm not convinced it's fixed, but I'll do more of what might have helped:

I raised the SI target voltage to 122 again. I had tried that before but didn't see a night & difference. I think I'll leave it that way because I was seeing about a 2 volt (of 120) drop at both bsmt and garage load panels with a moderate 2-4kw load on, like A/C, EV charger, induction range. There are two breakers between the SIs and each load panel. I might check those breakers resistances, I saw that discussed somewhere. I notice them buzzing a little bit, same with the GT roof array combiner box breakers. Maybe that's normal.

With the house loads on, I checked voltage at the microinverter combiners (way out at the garden, or the grid-tie array), while they were producing and I saw about 240-242. Both of these go to one of the load panels. So under a heavy load at the load panels, that voltage might be drooped from the load, and the micros' combiner is back above that close to the nominal voltage again. Under a light load in the house, there would be that much more, 4v at 240, rise out at the micros. We'll see if that's an issue but I doubt it. I see the grid here going to at least 122 or 123, and I see as low as 114 (with my house at end of line _pushing_ outflow back).

If voltage is a problem, it's transients, and the microinverters that I've been able to change the grid profiles on, or even the default grid profile, the voltages are quite a bit further out.

Last night I added more ferrite cores close to the microinverters on the original grid-tie array, and the battery microinverters. Maybe that's enough to help it. I will add more on the two circuits for the garden array and the 3 micros on garage roof. I wonder if it's better to put them out at the garden array, or in the load panel close to the SIs?

I removed the start thermistor and added a 521 kit start cap and relay to the A/C. I don't think that matters, but it was only $30 and easy, shouldn't hurt, until the start relay fails...

We'll see if it keeps working. Last summer it seemed fine, but at that time the 9 garage panels were DC coupled to the FLA battery. Now they go to the 48v lithium bank, and AC couple to the SIs (turn on the 12 microinverters) once that battery is full. So it's all AC coupled.

Thanks for the input!
 

Attachments

  • IMG_20220624_205727117.jpg
    IMG_20220624_205727117.jpg
    280.3 KB · Views: 2
  • IMG_20220625_160939683.jpg
    IMG_20220625_160939683.jpg
    279.3 KB · Views: 2
I had not looked at the logs recently until now, good idea. I ran everything off grid again today (it worked better today somehow...)

When it's ringing / oscillating / growling, it will set AC Frequency Out Of Range and Grid Instability. However, it will also say these normally, whenever the SIs raise the frequency to reduce AC couple power input.

There are also events today for AC Voltage Out Of Range - Phase 1 and 2, but I'm pretty sure these are when I was switching the load panels from off-grid inverters to grid or vice versa. The grid-tie system with it's automatic transfer switch also will set these.

Enphase M-series (as opposed to newer iQ-series) cares about L1-N and L2-N voltages. Since you already confirmed the L1-L2 voltage is fine then the ACVOOR events may indicate potential neutral issues.
 
OK, check line to neutral voltage.

It could be his DMM confirmed correct true RMS voltage, but inverter measured a different voltage due to distorted waveform.
 
Back
Top