Two units in parallel - grid power shown on GX web page is 'wrong'

DIY
, , ,
1

Hi guys, first time posting on here, I usually manage to sort things without too much trouble, but I seem to have an issue that has me stumped.

I have put it down as DIY, I am not an installer/professional, neither an expert - but I do have good energy management and electronics design experience.

I have used a MultiPlus-II 48/5000 for over a year without issue.

I wished to increase the charge rate on the batteries, and had the opportunity to get another MultiPlus-II 48/5000

All the obvious has been done, similar serial numbers, updated to have the same firmware etc. VE Bus setup complete so that one is master and the other slave. I have an external (Victron) CT around the meter tails (unchanged since running a single unit).

But I am seeing odd power readings on the Cerbo. This started as soon as I installed the second unit - perfect before hand.

Screenshot 2026-03-17 111900
Screenshot 2026-03-17 1119001087×654 41.5 KB

A ‘real’ grid meter actually shows grid power as zero - which is also the set point… so ‘correct’.

From the above, the 3570W is feeding nominally 2200W to the batteries, and the remaining nominally 1200W to the AC side.

The load is actually around 1200W at this point.

So, physically, the values add up, but the Cerbo is thinking 650W is going back to the grid when it is in fact going to the load.

As I say, it has a physical CT on the grid supply, so the reading should be right.

I am mostly sure that the AC supply and load are always split 50/50 in the display. This leads me to think there is some odd divide by two (due to the two units running in parallel) - though that isn’t proved yet.

Two easy questions;

When setting up, I find it odd that you have to install the ESS on both MultiPluses - I would have expected the master to transfer this - but no issue as long as my understanding is corrects. So, could someone confirm that this is the case, and ESS does need to be installed on both ‘manually’? It seems a receipe for disaster as you could set different battery sizes etc and really mess things up if you tried.

The second trivial question is, when setting up the first, obviously the external CT is selected. I have assumed that on the second, external CT should NOT be selected on the basis that IT doesn’t have one… but it is ambiguous as overall, the system does.

Any ideas would be most appreciated as I am not quite sure what to look at next.

Thanks

2

Each unit must be configured identically.
The master unit will control the assistant settings but it is needed on both - refer to the parallel setup training videos on the pro site.

The CT must be configured and enabled on the master, it will be the reference and it must be placed on a common section of wiring as is appropriate.
You also want the newer unit to be master.

Something is off in your config or install for this to happen.

I created nodered flows to chart individual performance in a parallel system. See here:

3

Hi Nick,

Thanks for the quick reply.

It is configued as you say - I just wanted to check that my interpretation of the docs was right and I wasn’t repeatedly reading it incorrectly.

ESS is installed on both as I believed correct - thanks for the confirmation.

Also, the CT is as you say on a common part of the wiring - actually around the meter tail before the CU and obviously then before the two Multipluses. Thanks for confirming that only the master has the external CT setting.

I will take a look at your flows a little later.

The newest unit is actually the slave at present…. but there isn’t much in it.

HQ2311 and HQ2346… I am assuming those are year and week - I would post the trailing letters, but apparently we shouldn’t post full serial numbers on here.

4

I meant only the CT on the master is used.
I would always configure units identically. An unnecessary setting will be overridden by the master anyway.

Do a test, disable the external CT for both, see if it behaves (assumes loads on output). Else enable on both.

5

I tried to debug by using VictronConnect, but frustratingly a few seconds after connecting, it puts the MultiPlus in to passthrough mode, so you can’t monitor what it is doing.

I managed to grab the screen on initial connect…

Victron
Victron799×630 53.7 KB

The actual grid supply at this point is nominally zero. The actual system load was most likely about 3kW (I can’t 100% prove that as I need to refer to different measurements in real-time and the Victron Connect values only exist for a few seconds before it put it in to passthrough).

Again, if my above assertion is correct, then again the displayed values match (the messed up) expectation of 3kW from the battery, zero from the grid, 3kW to load… which the MultiPlus thinks is half to grid, half to load.

It does however, get the Cerbo out of the picture for being in anyway ‘wrong’… the above is what the Multiplus ‘thinks’.

6

You can’t connect a mk3 and cerbo at the same time, removing the GX will result in passthrough.
One of the reasons I created the nodered visualisations.

Screenshot 2026-03-17 at 14.54.47
Screenshot 2026-03-17 at 14.54.47944×1300 65.8 KB

A negative AC in value is exporting. I would double check the CT and direction. Disable the CT altogether for testing.

7

Okay,

will make sure that the setting for external CT is ticked on both - though obviously only fitted on the master.

As I understand it, the temperature sense and voltage sense also only get used from the master - but I don’t have a problem there so all should be good.

Just one bit of extra info that might be relevant - this system is what is described as ‘grid parallel’ if I have my terminology correct.

i.e. the Victron system is connected only* via AC in. Hence needing the CT on the tails.

*Acually the always on AC out is used to feed some critical loads

8

Ah ok, that wasn’t clear but assumed some loads else you wouldn’t need a CT.
On the GX have you configured the load position in system setup, ESS etc?

9

Just checked, yes still set as Position of AC Loads = AC input only

Everything ‘looks’ right - there is obviously something unique to how I have it setup, otherwise it would be a known ‘mistake’ with an easy resolution.

I need to take a break from it now as I have other work to do… back on it in an hour or so.

I hoped to get it fixed before we enter Octopus’ Agile peak pricing at tea-time. Batteries will hopefully be charged since it is sunny - but who knows what it will do if it can’t work out its energy flow.

Thanks for your interest and help so far.

10

If you have critical loads on AC out it should be set to both. Not that it breaks anything but it has an effect on some visualisations.
Good luck

11

Ah, yes, you are right.

I confused myself slightly… the critical loads have just been wired, but aren’t actually turned on at the ‘always on DB’ yet.

So again, going forwards, I should change the Cerbo config… but it isn’t in play here with this problem yet.

12

More confused.

Just taken some AC current measurements with a Fluke clamp meter…

At the point between the CU and the isolator (taking all the load in/out to the whole Victron system)… about 5.5A which matches with a nominally 1100W load and zero drawn from the grid. So, that is as you would expect.

Measuring at each MultiPlus II, nominally 2.7A in each branch - ie. the two units are sharing the load, again that is as you would expect.

But on the display, again we see it claiming about 550W being exported, and 550W going to the load.

The units then decided not to charge the batteries - a ‘problem’ I think I do understand - on setup it wiped by battery ‘tuning’ and replaced with defaults, so I can address that. However, when not charging and not inverting, clearly the energy in should match energy out - it is effectively passthrough. And indeed it does show that correctly on the Cerbo. It showed nominally 1200W from the grid, zero to/from batteries and nominally 1200W to the load.

I thought I took a screenshot - but didn’t - so don’t have the proof to post… but it showed the charger as being in sustain, oddly even though the LED on the unit was saying bulk.

So, I think, this proves the current measurement is right (it’s wiring is unchanged since running for over a year anyway - but just in case I broke something whilst rewiring for the second unit).

Though I realise this is extremely unlikely, and I am the one to find it, but it is looking like a bug somewhere in the measurement system. If I knew which displayed values were physically measured, and which are derived, I might be able to narrow it down a bit more.

13

There are so many systems like this out in the wild, more every day.
If it was a bug we would know it by now.
You’re missing something.

14

Absolutely. The most likely fault here is me.

However, I have gone through everything I can think of by way of proof.

Just tried a test of setting the set point on the ESS to 3kW.

The physical grid meter (well the in-home display part) reads 1500W.

I am running an external system logging to MariaDB, it pulls the data from the registers in the Cerbo…. and again all the demand readings are around half that which I would expect. (Annoyingly, it uses that to forcast tomorrow’s demand, which it will now think is half as much as expected - but that is a trivial problem to fix)

Screenshot 2026-03-18 092844
Screenshot 2026-03-18 0928441142×704 42.3 KB

Image above is with a grid target of 3000W set in the Cerbo. The load should be around 3kW at this time… again, that displayed is about half.

For clarity, it isn’t simply that all the values are half… I think some are double… which almost certainly comes from a derived value being based on an incorrect measurement.

In theory, I could swap the units around, but would have to physically move them so that the CT wiring was on the right unit.

My present thought - and it is no more than that - is maybe an internal CT has failed. It seems to measure correctly from the external CT (proved when in passthrough), but just wondering if an internal measurement is ‘wrong’. It is however extremely unlikely to die just as I add a parallel unit.

I suppose I could prove that by setting the unit back up as a single.

15

Sometimes it pays to start clean. Reflash the units.
Change the master to the other unit.
Did you enable CT on both in veconfigure?

The best test is to have a critical load and just use the internal meters so you can eliminate the issue.

It is probably something obvious being missed. It is difficult to triage this over the internet when you can’t be hands on.

16

I did try the setting for external CT on both.. in fact it is like that now.

I tried internal - but then realised after the test, that I would have had to put the shorting link in the CT connector for the internal reference. It is obviously more than just selecting the tick box, I need that physical change.

Thinking now, I didn’t check the physical state of the CT plug on the new unit. To do that, I have to take the covers off again. I am on my own today and don’t want to be doing too much with covers off etc. in the cellar/plant room without some back-up in case I do something silly.

I think I am confident the CT is fine… based on the fact it hasn’t changed for over a year, and also that when in passthrough, the value displayed is accurate. The only thing I could have done, was trap or pull the CT cable when rewiring for the second unit.

I did a part re-install last night. I didn’t ‘create a new VE bus system’ but ‘modified and existing’.. removed the ESS assistants and started from that point.

I could, as you say, reset the units, and then setup again as a new VEbus system.

17

If you could, sharing all VE.Configure screenshots is very helpful.
Sometimes it even helps to provide a schema that shows how all is connected in and out the Multis and CT’s and so on.

18

Will get that done and posted later this evening.

19

I have deleted my ‘solution’ as further testing has revealed it is not completely correct.

I am working with Victron to get this solved, and will come back to this thread to update when the issue is properly sorted.