Parallel ESS units sporadically go out of balance

DIY
, , , , ,
1

Hi there,

Has anyone experienced a problem where parallel units within an ESS system sporadically become “unbalanced”?

The system can run for about a month at a time in a very stable manner, but about once per month, we notice the system fails to achieve the ESS grid setpoint.

Upon further investigation, it always comes down a very strange issue where much more power is flowing through one unit within a parallel set. This seems to throw off the ESS control loop.

Here’s a screenshot from the last episode of this behavior, showing (a) AC input power, (b) AC output power, and (c) inverter power, for all units. This was displayed via a node-red dashboard.

IMG_0114
IMG_01141481×1304 165 KB

Normally the system operates fine, and all 4 units are inverting at the same “inverter power” (as expected when ESS is set to “total of all phases”). I’ve not yet verified whether the imbalance within a phase is “real” via a clamp meter (because I forgot to check) — but the readings above do seem to indicate both phases are inverting at different power levels, which is not the correct behavior (and leads to unexpected grid import/export).

To resolve this imbalance, we flip off the AC breakers to/from both slave units, let the system run for a bit with heavy load — so all power is flowing through the masters — and then flip the breakers back on (a “live” fix). Rebooting the units sometimes resolves it too. It’s somewhat inconsistent which of these two fixes it.

Some details on the system:

  • Installed ~4 months ago
  • 4x Quattro 10k 120v units, in a 2x2 split-phase configuration
  • Firmware version v556
  • Cerbo on latest beta or stable firmware (behavior happens on both)
  • We followed the wiring unlimited manual very closely. All terminals were torqued within spec of the manual, identical wire lengths for units in parallel, of minimum gauge for our building code.
  • Our dealer ordered a dedicated batch of 4 units for our install (serials indicate they were manufactured within 1 week of each other).
  • 60kWh Discover AES battery bank with CAN-bus BMS
  • Grid code set to “other”
  • ESS grid setpoint of 0W, and multiphase regulation set to ‘total of all phases’
  • All loads are on the output side of the units, with ESS using the Quattro’s internal meters

We’ve tried the following, which didn’t help:

  • Flipping the master/slaves in software
  • Changing grid input connection from AC In 2, to AC In 1
  • Changing multiphase regulation to “individual phase”
  • Restoring configuration to default, and rebuilding the settings
  • Double checking the AC/DC connections are torqued and proper. For example, when this imbalanced behavior is occurring, we can switch to “inverter only” mode and everything will correctly operate in-balance in terms of AC Output and Inverter power. So this to me implies the AC output / DC connections are correct. This could still imply the AC input side is incorrect, though we’ve triple checked the torque on these (e.g., when switching from AC In 2 to AC In 1).

The fact that (a) the system can run stable for weeks at a time, and (b) a reboot / toggling a breaker resolves this implies to me there is probably a firmware issue somewhere (e.g., see this thread where someone else’s imbalance issues were resolved via a firmware update a couple years back).

I’ve asked my dealer for assistance, but would also appreciate the wisdom of the crowd here.

2

Just FYI, if you use firefox the gauges will render fine.

A parallel set shouldn’t run with one unit being grid disconnected, they function as one logical unit, it will drop both iirc, are you certain the config is right?

3

Are you certain that applies to slave units within a phase? In my case, the grid sensing seems to be done by the phase masters. It’s helpful to know either way.

4

If you use an external CT only the master will do sensing. It certainly isn’t supported to do that nor good for the relays which are supposed to switch as a group.

5

We aren’t using external CTs. LOM is set to Type B.

My understanding is in a multiphase setup, there is an overall “leader” unit and then there are masters for each phase.

If there are multiple units per phase (let’s call them slave units), those units are (literally) driven by the phase masters.

I’m not sure I understand how those slave units could do LOM detection, since LOM detection depends on the phase master (which sets the frequency) attempting to push / pull the frequency.

Happy to be mistaken as it’d provide some sort of path to explore.

6

Been chatting internally with the group, my understanding was incorrect, it applies only to 3-phase and switch as a group enabled.
It is not recommended to disconnect either IN nor OUT on one unit. AC OUT may actually be more problematic and I would have expected that the master would drop its supply if a slave lost its grid connection. I can’t test on mine at the moment.

Imbalances tend to be fixed and easily visible as load increases.
That is why I wrote that flow that you have above, to test my own at different loads.

When strange behaviour happens that resolves by restarting things, it can be different reasons, some times a wiring issue that was overlooked - as seems to be the case on a different topic, recently.
Balance is typically determined by wiring, and given how intermittent this is, I can only speculate there is something on the mains supply for that phase.

What gauge wiring are you using and what is the length?
Anything different on that phase?

Its always a challenge trying to diagnose this stuff remotely, sight unseen.

7

Thanks for looking into the LOM topic - helpful information.

On the AC input/output side, we’re using 11.5 feet of #2 awg wiring for L2 units, and 15.5 feet of #2 for L1. These tails connect to 100A breakers, which then feed into a power distribution block going to 200A #4/0 feeders (from grid supply / to building loads).

For the DC side, each inverter is fed by similar length #4/0 cabling (one pos / one neg), connected to 250A breakers in a 1000A battery combiner box.

All cabling was measured to within 1 inch, terminated with copper lugs and a 10+ ton crimper, and connections torqued. My distributor has reviewed the system line diagram and photos of the install itself, including the configuration file, and couldn’t find issues.

Nothing is different for the specific phase in the example above, and it’s not consistent to that phase if I recall correctly.

I’m convinced it’s not related to the wiring for 4 reasons:

  1. If the system is showing that imbalanced behavior while grid connected, and we change the mode from “On” to “Inverter only”, the whole system balances itself out. Meaning, “inverter” and “ac output” power become balanced between master and slave units. To me, this implies the AC output and DC connections are solid, given no physical interaction happened with the system (e.g., “knocking” a wire).

  2. If we change the system mode to “Keep batteries charged”, the system again balances itself out on the “AC input” and “inverter power” side. This to me implies the AC input wiring is solid, and again affirms the DC wiring was solid, for similar reasons as above.

  3. When reverting to “On” mode, the issue returns, which can then be resolved by either rebooting the units or togging the slave breakers.

  4. The “inverter power” in the photo above is not equal between L1 and L2, but equal between units within each phase. When ESS in “total of all phases”, all units in the system should be inverting at the same power.

As someone with a computer science background, it almost feels as if the control loop is losing sync between units somehow (or “locks” at the wrong value).

Let me know if you disagree with my logic… I’ve been battling this for months. Please let me know if you have any other debugging steps or if there’s something I can check next time we see this issue.

8

Balance is purely controlled by resistance, provided by cabling and the inverters themselves.
A conventional balancing issue is consistent, but can increase with usage and often the best way to improve it is to add more resistance, so longer, thinner cabling.
I did this quite successfully with my own, home, system.

In your case it does not fit this profile, and the two units don’t operate independently. From an operational perspective it is one system.
If this behaviour is swapping phases then that is even more confusing.
With phase regulation the system will try balance to a net zero, so that can move around depending on load, but that should be balanced across inverters as well.

We haven’t seen anything like this reported before, and the code is common, as are multi-phase systems.
Usually it is a devil hiding in the detail.

Can you upload your config here (the whole configurator file, you can pull it remotely) not just a veconfigure from one of the inverters.
There is no personal data in it.

9

Joining in on this from the distributor side, A-P and I have been working through this in detail for awhile now.

@A-P , there have been a couple screenshot examples of the Leader/Follower in a phase operating opposite as well, haven’t there? I think I’m recalling one example with the leader pulling from the grid while the follower sent to the grid, though I can’t find that one in our coorespondence. Same behavior as an imbalance, but notable because of the opposing operational state. If you have that dashboard screenshot handy could you upload it for the thread as well.

10

Are the units all the same age?
I have seen that behaviour before with mismatched systems, but in that case swapping the master role to the youngest system resolved it.
This was constant behaviour, not intermittent, as the disparity in either wiring or system age is a constant.

Given the complexity here, as the supplier, you should consider involving L3 support, which you should have access to.
It is just too difficult to assist remotely.

You are welcome to upload the config file.

11

Thanks for digging in here @nickdb Truly appreciate it.

My understanding is that when ESS is set to “total of all phases”, all 4 units (across all phase) will invert at the same power level (per the docs). So this will usually result in some grid export on one phase, and some grid import on another phase – but overall net zero sum. I’m correctly seeing this behavior most of the time – when the system is behaving properly.

However, what is weird in the example I posted above: The two units on L1 are correctly inverting at 2.7kW each, and but the two units on L2 are “under” inverting at 2.0kW each (rather that 2.7kW).

Even with some resistance issue, I would expect the ESS control loop to still “force” the units on L2 to invert at 2.7kW each, given all four units should be inverting at the same power level when ESS is set to “total of all phases”.

This results in a 1.4kW deficiency on L2 (0.7kW * 2), which is being made up via “AC In” (grid).

The readings via vebus are also showing the L2 master importing ~0W, while the L2 slave is importing 1.3kW (a different type of imbalance). Though I don’t believe this is a “real” imbalance (due some sort of resistance) for the reasons I previously posted.

I will admit we have not current clamped the individual units when this issue occurs, to verify if this imbalance is “real”. I will do so next time. I’m assuming we’ll see the grid draw balanced between the two units, on the phase which is under-inverting.

That said, I’m not sure it would explain the different “inverter power” readings between the two phases, which still leads me to believe in a software-related issue.

Let me know if you or the team think otherwise.

The units appear to be the same age:
|Phase 1 Master|2025-06-10|HQ2337*****||
|Phase 2 Master|2025-06-10|HQ2337*****||
|Phase 1 Parallel Unit 1|2025-06-10|HQ2338*****||
|Phase 2 Parallel Unit 1|2025-06-10|HQ2338*****|

Please see attached my config file as well.

Thanks again.

I will also post a few screenshots shortly showing the behavior that @rloessberg mentioned as well (thanks Reid) - I have a couple, but slightly less detail than the dashboard posted above given it was early in my debugging attempts.

config.rvms (9.2 KB)

12

One observation from the team, you appear to have mixed the slightly older and slightly newer units across phases, I would put the xx38’s on one phase as a parallel pair, and the xx37’s on the other. Maybe there is a minor difference between them causing issues.
Beyond that can’t say I can see anything obvious. I suggest the best course of action is to have your supplier escalate to L3 support.