Solar Edge Modbus limiting intermittent

Interesting one this — anyone else running AC-coupled solar with Modbus frequency shifting and seeing the export limiting drop in and out unpredictably?

I've got a much simpler setup on my narrowboat (Victron Multiplus-II 3000 with a small Renogy array), but I've been helping a mate on his static off-grid setup sort out a three-phase arrangement with a SolarEdge inverter talking to his Victron system via Modbus TCP. The frequency-shift/ESS limiting works fine for days, then randomly the SolarEdge seems to ignore the curtailment signal and just hammers the batteries until the BMS trips.

A few things we've checked so far:

• Network stability — Modbus TCP relies on a solid LAN connection; one dodgy switch and it all falls apart
• SolarEdge firmware — there were some known issues with earlier versions handling Modbus commands properly under fluctuating load conditions
• Polling interval — the Venus OS default might not be aggressive enough to keep SolarEdge happy

My suspicion is it's a firmware/timing handshake issue on the SolarEdge side rather than anything the Victron is doing wrong, but I'm not certain.

Has anyone here run a similar AC-coupled SolarEdge + Victron multi-phase setup and cracked this particular gremlin? Specifically curious whether switching to frequency shifting instead of Modbus makes it more reliable, even if you lose some of the fine-grained control.

Would also be interested whether anyone's had SolarEdge UK support actually engage meaningfully on off-grid Modbus configs — they tend to assume you're grid-tied and get a bit lost otherwise.

@SmartSolar_Master Modbus TCP timeouts are the usual culprit — if the SolarEdge inverter misses a few polls it'll revert to default behaviour faster than you can say "why is my battery full and my neighbour's grid connection on fire." Check your network switch isn't dropping packets, set the Modbus timeout to maximum, and make sure your Cerbo GX (assuming Victron) is polling at no more than 1 query per second. Also worth checking the SolarEdge inverter firmware — they've quietly fixed frequency-shift response bugs in a couple of recent releases. On my shepherd's hut setup the AC-coupling was rock solid once I moved the Cerbo onto a wired ethernet connection rather than WiFi.

@MarineGeoff is right about the timeouts, but worth checking the poll interval too — SolarEdge defaults can be aggressive and if your network has any jitter the inverter will revert to its last known state mid-cycle.

Two things I'd add:

1. Modbus register 61445 — check you're writing to the correct export limit register and that it's not being overwritten by the SolarEdge cloud service simultaneously. Seen this bite people badly.
2. If you're using a Victron CCGX or Cerbo as the Modbus master, verify the VRM connection isn't stealing priority on the same TCP stack — there's a known timing conflict.

On my shepherd's hut system I moved to frequency-shift limiting via the inverter's own grid code settings rather than active Modbus polling. Removed a whole layer of failure modes entirely. Might be worth considering if your setup allows it.

@SmartSolar_Master bit of a tangent from the SolarEdge discussion, but your narrowboat mention caught my eye — I spent three years wrestling with AC coupling on my own boat before switching to a simpler DC-coupled arrangement with a Victron MultiPlus-II.

The frequency shifting approach always felt fragile to me in a marine environment. Vibration, temperature swings, damp connectors — any of those can introduce just enough latency into your comms chain to trigger exactly what you're describing.

Before assuming it's purely a Modbus timeout issue as @MarineGeoff suggests, I'd physically inspect every ethernet or RS485 connection in the loop. On my boat I found a partially-corroded RJ45 that was intermittently dropping frames — looked perfectly fine visually but caused maddening behaviour that took weeks to diagnose.

Crimp quality matters enormously in that environment.

@SmartSolar_Master worth double-checking your Modbus register mapping too — SolarEdge sometimes silently ignores write commands if the inverter firmware version doesn't match what your controller expects. Victron's ESS assistant handles this slightly differently to raw Modbus implementations, so if you're using a GX device in the loop there's a specific setting under Grid Meter that can cause phantom dropouts even when the TCP connection looks healthy.

Also, are you on a fixed IP for the SolarEdge unit? On a narrowboat the network can get a bit sketchy — any DHCP lease renewal mid-poll will cause exactly the intermittent behaviour you're describing. Static IP sorted a similar issue for me.

What GX firmware are you running? There were some known Modbus stability improvements around v3.x that might be relevant here.