VPP idle mode

[ridax67]

I know this has been discussed in another PR but can't find which.

There was a long period of idle during night but battery was not empty, because there as a plan to cover more expensive morning hours with the remaining energy, or there was much emergy left for morning dump.

We found that battery first 1 is the state that best holds battery level static which it is.

But it has the drawback that for those cases that idle is used because there is almost no energy left (soc min) the battery and bms is not going to sleep.

This could be fixed by either using load first in those situations instead (idle at soc min) or use power 0 when soc is at soc min. Both these modes will let the bms sleep when long periods of idle with no energy.

I have tested to forced power to 0 when soc is at min, and I think it is working nicely so recommend this, although not really incorrect as it is.

[johanzander]

Thanks — you've found a real gap, and your description matches the code exactly. The PR you were thinking of is #466 (see also #488 / #517).

What happens today

Whenever the plan says IDLE, we put the inverter in battery first — an actively-held command meaning don't touch the battery, serve the house from grid/solar instead.

That is correct when the battery is deliberately holding energy back for the morning peak. It is pointless when the battery is empty: there is nothing left to protect, but we still keep the inverter under active remote control all night, so nothing gets to sleep. Worse, because remote control stays enabled we re-send that command every 15 minutes for the entire idle stretch, refreshing the inverter's fallback timer each time — so it is never released.

Your two options differ in one way that matters: surplus solar

There are three states available while idling:

Mode	Battery	Surplus solar
Battery first (today)	untouched	still trickles into the battery
Grid first (your "power 0")	held against charging	bypasses the battery, goes to grid
Load first (your other option)	inverter's own self-use	still absorbed into the battery

The optimizer's internal model of IDLE explicitly assumes surplus solar does get absorbed. So forcing power to 0 quietly breaks that assumption: any idle-at-empty period with sun would export solar the plan expected to bank.

You wouldn't have seen this — you tested overnight, where there is no sun and the two modes are indistinguishable.

I measured how common idle-at-empty actually is across our 37-day test corpus: it occurs on 24 of the 37 days, and one winter day is idle-at-empty for all 24 hours. So it is not a corner case, and several of those days do have sun.

Recommendation: load first (release control) at minimum SoC

1. It changes nothing about energy flows. At minimum SoC there is no energy to discharge, and solar absorption keeps working exactly as before — behaviourally identical to today in every case, so it cannot regress Question: How is IDLE used? #466.
2. It is the only option that actually releases the inverter. "Power 0" keeps remote control switched on, so we would still re-send a command every 15 minutes all night. Load first stops the writes entirely and lets the inverter's own dead-man timer lapse — the inverter is genuinely handed back, not just commanded to zero. That is what lets the BMS sleep.
3. Nothing can drain below the floor, because we already write your configured minimum SoC into the inverter's own hardware cut-off register at startup. That is a firmware-level guard, independent of whatever we command.

The trigger is the same either way: only when the plan says IDLE and the battery is actually at its minimum. Above the floor nothing changes — battery first still holds the energy for the morning.

So: your instinct was right, and your first suggestion (load first) turns out to be the safer of the two you listed. Your assessment that it is "not really incorrect as it is" is also right — this is a BMS-sleep improvement, not an optimizer correctness bug.

One thing worth flagging: both the battery-first and grid-first VPP semantics are still marked not yet real-hardware-validated in our docs, so once this lands it would be very helpful if you could confirm on your inverter that (a) the BMS does go to sleep during a long idle-at-empty stretch, and (b) the battery does not discharge below your minimum SoC.