Oscillating on a delta grid connection (3x230 without neutral)

Got it stabile and with all 3 phases indicated.

What I’ve done is below. Also somehow in line with above document.

So, I’ve started from the principle that L2 is reference and set PV inverter and Victron on same phases. Also, aligned with Victron recommandations: L1 to L, L2 to N. (Yes, if your cables are not colored as per standard (in my case also not) I took first wire coming from official energy meter as base of being L1.)
Then I’ve checked my electricity incoming and defined the rotation. I needed to arrange the incoming lines. With this, also the meter connected to Victron correctly reported L1-L2-L3. (Before the work, was L1-L3-L2).

Then, took time and calm, I’ve checked the phases used by PV inverter and also by Victron. And here is the trick. I’ve made them to be the same for both units.

Then, I’ve started the PV and let him start completely. (~10min)

Then, started the batteries and Victron.

Set reading of meter for all 3 phases: At Integration set the meter as Multiphase. PV inverter linked to phase L1, as is in real. At ESS - Multiphase regulation. (MP-2 would report as alarm Phase rotation unknown. Probably this one is ok, while I have single MP in the net.)

I could not believe it. It is running for several hours already and it does stay nice. Also now is balancing full power.

Conclusion: Instability in 3x230 without neutral would come from the fact that PV and batteries are actually fighting each other, as being on different connections. Check and correct the connections. Make them exactly same. Start one unit at a time. Happy days!

PS. A resistor of 120ohm on EM540 side might help for stability.
For now, settings of EM540 are reset to default each time modbus protocol is restarted (restarting Multiplus or reconnecting USB to Multiplus). I am in discussion with Carlo Gavazzi how to solve.