Dual Sol-Ark 15k Load Sharing Issue

[NullDev]

First, to describe the setup, I have two Sol-Ark 15k inverters set as Primary/Secondary. The 200A grid is supplied to both Sol-Arks via a PDB. Likewise, the load side is combined via a PDB and sent to the 200A main panel in the house (whole house backup). We have 20kWh of Li-Ion batteries paralleled and tied to both Sol-Ark units. Lastly, each Sol-Ark has a string of 8 panels on MPPT1 and a string of 9 panels on MPPT2 for a total of 4 individual strings. All the strings are ground mounted away from anything that might shade them.

I have Solar Assistant gathering data which is fed to Home Assistant via MQTT.

Now the question, I have noticed at various times through the day that the 2 systems do not seem to be equally sharing the load to the house. In some cases, one inverter will actually show a negative wattage. I initially thought that perhaps my Home Assistant was corrupting the data somehow, but I verified the numbers in Solar Assistant and then also with the MySol-Ark website and they all agree.

I've attached a screen grab showing one instance where the house load was around 850 watts. One Sol-Ark was providing 1246 watts to the house and the other showed -355 watts.

This doesn't seem "normal" to me. Can someone confirm if this is expected? If not, any clues where I should start looking to resolve the issue?

Configuration:

• Work Mode = Limited Power to Load
• Solar Sell = Enabled
• Energy Pattern = Load First
• TOU = Disabled
• Max Sell Power = 9000w

 

[NullDev]

So, no one here has ever put 2 Sol-Ark 15k's in parallel and checked how they share a common load?

 

[HighTechLab]

Any photos of the install? How (if) are the CT's installed? Is this while the grid is active, or in inverting mode?

 

[wpns]

One Sol-Ark was providing 1246 watts to the house and the other showed -355 watts.

This doesn't seem "normal" to me.

That’s really wierd, what would the second inverter do with the power? Even if the measurement was L1 versus L2 it makes no sense.

 

[Solar Guppy]

That’s really wierd, what would the second inverter do with the power? Even if the measurement was L1 versus L2 it makes no sense.

One could be charging battery the other selling to the grid?

My Solis S6 does this occasionally, its software controlled to try and maintain a balance between the two and with load changes or battery charging power or solar differences energy can be pulled from the other inverter from the backup port instead of the grid.

 

[NullDev]

Any photos of the install? How (if) are the CT's installed? Is this while the grid is active, or in inverting mode?

There are no external CTs installed, so I assume the on-board CTs are being used. The grid was always active.

 

[HighTechLab]

Are the cables to and from each inverter approximately the same length? If the transfer switch relays are engaged to grid, then grid load is only balanced / equalized by the resistance of the wires.

Still no pics of the install?

 

[NullDev]

One could be charging battery the other selling to the grid?

My Solis S6 does this occasionally, its software controlled to try and maintain a balance between the two and with load changes or battery charging power or solar differences energy can be pulled from the other inverter from the backup port instead of the grid.

I wondered if it wasn't something like that. However, the batteries were full and most of the generated solar was getting pushed out to the grid. Unfortunately, I didn't look to see how the 2 units were sharing the grid export balance. Since both Sol-Arks are fed with the same panel configuration and none of them are shaded, You'd think it would be easy for the Sol-Arks to balance.

Maybe there is more going on here than is obvious?

 

[NullDev]

Are the cables to and from each inverter approximately the same length? If the transfer switch relays are engaged to grid, then grid load is only balanced / equalized by the resistance of the wires.

Still no pics of the install?

I'll dig through and see if I can find a good pic. The 2 Sol-Arks are mounted with a 12"x12" trough below which has the power distribution blocks. There is a foot or less of difference in the 4/0 lengths. As a side note, the load sharing wanders back and forth between the 2 Sol-Arks. It's not always the same one doing the bulk of the work.

 

[Solar Guppy]

Maybe there is more going on here than is obvious?

Being software controlled it is impossible to always have perfect balance between parallel units and aside from an OCD concern, it has zero impact on system performance.

 

[NullDev]

Being software controlled it is impossible to always have perfect balance between parallel units and aside from an OCD concern, it has zero impact on system performance.

I completely understand that it wouldn't be 100% balanced. However, when I saw one was supplying 1246w and the other was consuming 355w, I got concerned as that didn't seem "normal". I didn't want to pay for 2 Sol-Arks only to be limited to the capability of 1 because they can't load-balance properly.

 

[NullDev]

Are the cables to and from each inverter approximately the same length? If the transfer switch relays are engaged to grid, then grid load is only balanced / equalized by the resistance of the wires.

Still no pics of the install?

Here's a pic earlier in the build process where it's fairly easy to trace the 4/0 runs. The block on the left is the load side and the block to the right is the grid side.

 

[Nobodybusiness]

Here's a pic earlier in the build process where it's fairly easy to trace the 4/0 runs. The block on the left is the load side and the block to the right is the grid side.

View attachment 267427

I have 2 Sol-Ark 15s in parallel just like you.

My load power is exactly 50% on each one of total load.

Look on the inverters themselves.

Looks like Home Assistant is reading load power non-essential registers instead of load power

 

[Quattrohead]

Yeah I would definitely look at calculation errors as opposed to actual problems, if you have a current clamp meter you can verify all readings.
I have paralleled EG4 18K PV's and they share the load very well together but you can definitely see a discrepancy at very low power consumption, occasionally one will be producing almost no output at all and the other one will be covering the 2 to 300 watts house consumption.
I have never seen a negative consumption.

 

[Quattrohead]

By the way this is all using Solar Assistant feeding data to Home Assistant.

 

[NullDev]

I have 2 Sol-Ark 15s in parallel just like you.

My load power is exactly 50% on each one of total load.

Look on the inverters themselves.

Looks like Home Assistant is reading load power non-essential registers instead of load power

I initially thought that as well, but I checked the Sol-Ark display and they were indeed showing 1246 watts to the house on one and -355 watts to the house on another. It was not the non-essential registers. I have an Emporia Vue on the main panel in the house, and the net power received from the Sol-Arks was confirmed at around 850 watts, so the math works out.

I just don't understand why one Sol-Ark would be actively consuming power on the load side.

Unfortunately, I don't have a current clamp meter to verify the Sol-Ark readings.

Not sure whether this matters or not, but I have not installed any CTs. Since I'm not using Limited Power To Home, all current measurements are being done by the onboard CTs.

 

[NullDev]

By the way this is all using Solar Assistant feeding data to Home Assistant.

My data gathering setup is similar with Solar Assistant feeding Home Assistant. All of the data sources match between MySolArk, Solar Assistant, Home Assistant, and the GUI on the front of the units.

I can understand the load sharing being off by a few percent under load or a little wonky during low loading. I find it hard to understand a wildly negative load number under a semi-reasonable load.

 

[wpns]

If all the settings are in sync (EG4 has the option of copying all the settings from Primary to Secondary, dunno about SA) then I’d call their tech support. Feels like one inverter is trying to charge the batteries while the other runs the house load and the additional load of the battery charging function. #PerpetualMotion.

 

[Quattrohead]

This is what it sounds like to me too

 

[Nobodybusiness]

If all the settings are in sync (EG4 has the option of copying all the settings from Primary to Secondary, dunno about SA) then I’d call their tech support. Feels like one inverter is trying to charge the batteries while the other runs the house load and the additional load of the battery charging function. #PerpetualMotion.

If it’s setup as a True Master Slave that can’t happen.
The settings are the same across the pair.

Sound like need to call Support and see what they say.

 

[robby]

I also suspect it's a problem with the Master Slave settings.

 

[NullDev]

I also suspect it's a problem with the Master Slave settings.

I feel semi-confident it isn't a master/slave configuration issue. I'm able to make changes on the master and they automatically propagate to the slave.

 

[NullDev]

If it’s setup as a True Master Slave that can’t happen.
The settings are the same across the pair.

Sound like need to call Support and see what they say.

I sent an email to Sol-Ark's support team on Jan 1st with as much supporting documentation as I could gather. I haven't heard a response yet, but I will report back once I do.

 

[Nobodybusiness]

I sent an email to Sol-Ark's support team on Jan 1st with as much supporting documentation as I could gather. I haven't heard a response yet, but I will report back once I do.

You probably going to need to call them.

Email says to them “when you get around to it.”
Phone call says stuff is broke and need help now.

We used to have a guy on here that would look at stuff but I guess he moved on.

 

[NullDev]

You probably going to need to call them.

Email says to them “when you get around to it.”
Phone call says stuff is broke and need help now.

We used to have a guy on here that would look at stuff but I guess he moved on.

Judging by their response time, you may be right. I figured with all the screen image data, it would make more sense to do it by email.