Feed-in limit (MaxFeedInPower) not respected in "Total of all phases" mode with DC-coupled PV

Been scratching my head over something similar with my setup and wondering if anyone else has hit this wall.

Running a three-phase Multiplus-II arrangement (48V) with a Cerbo GX and a couple of Victron MPPTs for DC-coupled solar. Got the MaxFeedInPower set in VEConfig but on a decent sunny day it just... ignores it. The inverters happily push well over the limit back to the grid, especially when the Pylontech stack is already full.

From what I can tell, the issue surfaces specifically when you're in "Total of all phases" feed-in mode rather than per-phase limiting. Seems like the MPPT output doesn't get properly throttled in that scenario — the ESS assistant appears to lose track of what the DC side is actually contributing.

Tried dropping the feed-in limit right down to test it, same result. Grid meter is definitely reading correctly.

Few questions for anyone who's dealt with this:

• Did you resolve it by switching to per-phase limiting instead?
• Is there a known firmware version where this was patched? (I'm on latest but "latest" doesn't always mean "fixed")
• Anyone managed to get Victron support to acknowledge it as a bug rather than a config issue?

Bit frustrating because DNO export limits are a real thing here in the UK and I don't want to risk breaching my G98/G99 agreement over a software quirk.

Would be good to know if this is widespread or just my particular combination of kit.

@ShortCircuit DC-coupled PV just does whatever it likes when the sun's out — my static van's single-phase Multiplus-II has the same selective hearing when the MPPT decides grid limits are more of a suggestion than a rule. 🤦

@ShortCircuit this is a known limitation worth understanding properly. The MaxFeedInPower parameter in the ESS assistant operates on the AC output side — it throttles the inverter's contribution. But DC-coupled PV (MPPT directly on the DC bus) bypasses that control path entirely; the Cerbo can see the production via VE.Direct but can't instruct the MPPT to curtail fast enough to respect per-phase or aggregate feed-in limits reliably.

In "Total of all phases" mode specifically, the ESS algorithm averages across phases before acting, which compounds the lag.

What's actually enforced is the inverter output, not total DC generation. Victron's own CCGX/Cerbo documentation is quietly honest about this if you dig into the ESS design guide (section on DC-coupled systems).

My workaround on the shepherd's hut system was enabling DVCC with a tightly set SVS — doesn't fully solve it, but reduces overshoot significantly.

@ShortCircuit yes, hit exactly this with my setup last autumn. Worth checking your DVCC settings if you haven't already — specifically whether "Limit charge current" is properly configured, as that can interact oddly with feed-in behaviour on DC-coupled systems. The Cerbo essentially can't throttle a MPPT the same way it can control an inverter output, so there's an inherent lag before it reacts to grid export figures.

One workaround I found helpful was setting a more conservative MaxFeedInPower value than you actually want, to account for that overshoot. Not elegant, but it kept my DNO happy while I waited for a firmware update to address things properly. What firmware version are you running on the Cerbo? Some of the earlier 3.x builds had particularly erratic behaviour with this.

@BordersExplorer that's useful context — do you know if this limitation is documented anywhere in the ESS design guide? I've been through the Victron docs and can't find a clear statement that MaxFeedInPower simply won't constrain DC-coupled sources the same way it does AC.

On my narrowboat I only run single-phase so I've not hit the "total of all phases" wrinkle specifically, but I did notice my MPPT would happily push past what I'd set as the feed-in cap during peak solar. Eventually realised the Cerbo was calculating the curtailment headroom incorrectly because my DC loads weren't being factored in properly.

Is the three-phase "total" mode actually summing correctly across all phases before applying the cap, or is it applying the limit per-phase and then summing — because those would give very different results depending on load imbalance?