Grid export occurring despite PreventFeedback=1 and MaxFeedInPower=0 — 3-phase MultiPlus-II system

Been having a similar head-scratcher with my 3-phase MultiPlus-II setup and thought I'd throw it open here.

Despite having PreventFeedback set to 1 and MaxFeedInPower at zero across all three phases, I'm still seeing the odd blip of export registering on my DNO meter. Nothing dramatic — we're talking maybe 50-80W peaks — but it's enough to make me nervous given my G98 notification status.

My suspicion is it's a response latency issue rather than a misconfiguration. The ESS assistant reacts to the grid meter reading, but there's an inherent delay between measurement, processing, and the inverter actually pulling back. If a load drops suddenly (washing machine finishing its cycle, that sort of thing), the system briefly overshoots before it corrects.

A few things I've tried:

• Tightening the MaxInverterPower per phase
• Setting a small negative offset on GridSetpoint (around -50W) to keep a buffer
• Checking the CT clamp positioning on the grid meter — turned out mine was slightly off

The negative GridSetpoint trick seems to have reduced incidents noticeably, essentially telling the system to target a slight import rather than true zero.

Curious whether anyone else running Fogstar or other large LiFePO4 banks has seen this — I wonder if the faster discharge response of lithium compared to AGM actually makes the overshoot problem worse in some configurations.

Has anyone found a definitive fix, or is some tiny residual export just unavoidable with this topology? Would love to know if there's a VEConfigure setting I'm missing.

@LindaJones69 worth checking whether your ESS Assistant version matches across all three inverters — a mismatch can cause one phase to behave independently during the assistant's feedback-prevention logic. Also, PreventFeedback=1 alone isn't sufficient in some firmware versions; you need to confirm FeedInOvervoltage and the ESS grid setpoint are correctly configured in Venus OS, not just the individual unit settings.

On my Vivaro-based setup I caught similar phantom export because the DVCC current limiter wasn't propagating properly to the phase with the highest PV contribution.

Check your VRM portal → Advanced → ESS graphs — if one phase is consistently the exporting culprit, that narrows it down considerably. Also, which Venus OS version are you running? There were known ESS bugs around feed-in control prior to v3.10.

@VivaroWanderer makes a fair point on the ESS Assistant versioning, but there's another culprit I've seen bite people on 3-phase setups specifically: the L1 master unit is the one that actually enforces the feed-in prevention logic, so if your grid meter reading is arriving late or with comms errors, the master can momentarily "lose sight" of what the other phases are doing and briefly allow export before correcting itself.

Check your CCGX/Cerbo GX logs for any grid meter timeout events around the timestamps when export occurred. Also worth verifying your ESS minimum SOC isn't fighting against the feed-in prevention — counterintuitively, a very low SOC floor can trigger some odd compensation behaviour.

What grid meter are you using? Some of the cheaper RS485 units have timing issues that interact badly with Victron's control loop on 3-phase configs.

@LindaJones69 classic three-phase trap — ESS calculates feed-in based on total combined phase power, so if L1 is exporting 500W whilst L2 and L3 are consuming, the system considers net export zero even though your DNO meter sees real electrons leaving on L1; fix is enabling phase compensation in ESS settings or you'll be donating free electricity to the grid forever like some sort of renewable charity.

@SussexSolar nailed the phase balancing issue — worth adding that this catches people out even more when one phase has a heavy load pulling it negative while another's exporting. The net calc looks fine but L2 or L3 is quietly pushing to grid.

Check your ESS minimum system power setting too. Had this exact thing on my garden office setup — Victron Cerbo GX was showing zero feed-in but the meter at the DNO side disagreed. Turned out MinimumSocLimit interactions were causing brief export bursts during bulk-to-absorption transitions.

Pull the VRM advanced graphs and overlay all three phase outputs against your grid meter reading — you'll spot the pattern pretty quickly. 15-minute averages hide it, switch to 1-minute resolution.