fix(plan): don't offer Freeze Export for inverters that always charge from PV surplus

[chalfontchubby]

Summary

Fixes #4207. Supersedes #4425 (closed separately - see comment there).

Some inverters (e.g. FoxESS "Feed-in First") can't be commanded into a state where battery charging is disabled at all - they keep charging from any available surplus PV regardless of the mode Predbat has requested. That means Freeze Export (hold the battery flat, export the rest) can never actually be achieved on that hardware. Predbat's plan previously assumed freeze always holds SoC flat, so on this class of inverter it predicted "no charging" while the inverter actually kept charging from the surplus - a plan/reality mismatch.

#4425 tried to fix this by teaching the prediction model to simulate PV-charging during freeze too. That correctly modelled the physical behaviour, but as raised there: once applied, freeze (99%) and idle (100%) run through identical code for these inverters, so "Freeze Export" becomes a pure relabelling with zero distinguishing effect from doing nothing - the plan would keep showing a FrzExp slot that isn't actually achieving a freeze.

This PR takes the alternative raised in that discussion instead: rather than patch the prediction model to simulate a distinction that doesn't exist in reality, stop the optimiser offering/selecting freeze for these inverters in the first place.

Revision history on this PR, for anyone following the thread: this was originally gated on the existing inverter_can_charge_during_export flag, reusing it rather than adding something new. @springfall2008 correctly pointed out that's wrong - that flag is specifically about PV exceeding the inverter/export limit during active Force Export (see its own docs entry), not about whether Freeze Export can hold the battery flat at all. Two genuinely different questions. This is now gated on a new, dedicated inverter_can_freeze_export setting instead (apps.yaml only, defaults to true - most inverters genuinely can freeze, so "assume it works, let affected users opt out" is still the right default, it just needed to be its own flag).

• optimise_export(): allow_freeze is forced off when inverter_can_freeze_export is false, alongside the existing set_export_freeze check.
• optimise_solar() ("Export more solar"): sets export limits to 99.0 (freeze) directly, bypassing optimise_export()'s search entirely, so it needs the same gate independently or it reintroduces the bug through the back door.
• No prediction.py or kernel changes - this is purely a search-space restriction, not a modelling change, so no parity bump or binary rebuild needed.

Consequence worth being upfront about: this is opt-out, not opt-in - Freeze Export keeps working for everyone by default, and only stops being offered for someone who's explicitly set inverter_can_freeze_export: false because their specific inverter genuinely can't achieve it.

Test plan

• Regression test in test_optimise_solar.py confirming freeze is not offered when inverter_can_freeze_export=False, and unaffected (still offered) otherwise
• Existing test_optimise_solar.py scenarios pass their original intent unchanged
• ./run_all --quick passes
• ./run_pre_commit passes

🤖 Generated with Claude Code

[daviessm]

Does this mean that for inverters with inverter_can_charge_during_export=true, "feed-in first" mode won't be used at all and the only way to get the system to export before the battery is full would be to set force discharge?

[springfall2008]

I'm not sure this seems right, freeze export disables battery charging allowing all solar to be exported.

Can charge during export just means can the battery charge if the solar exceeds the capability of the inverter to export, this is different.

[chalfontchubby]

I wondered about a new switch - battery can’t not charge while exporting,Freeze_export_supported, something like that    Sent from Yahoo Mail for iPhone On Sunday, August 16, 2026, 6:37 pm, Trefor Southwell ***@***.***> wrote: springfall2008 left a comment (springfall2008/batpred#4538) I'm not sure this seems right, freeze export disables battery charging allowing all solar to be exported. Can charge during export just means can the battery charge if the solar exceeds the capability of the inverter to export, this is different. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***>

[chalfontchubby]

@springfall2008 You're right, thanks for catching that - inverter_can_charge_during_export is specifically about PV exceeding the inverter/export limit during active Force Export, not about whether Freeze Export can hold the battery flat at all. Conflating those was a mistake.

Pushed a fix using a new, dedicated inverter_can_freeze_export flag instead (default true, so nothing changes for anyone unless they explicitly set it false) - PR description updated to match.

[chalfontchubby]

@daviessm To answer directly: no, this shouldn't take away "feed-in first"/pre-emptive export for you. The gate now checks a new inverter_can_freeze_export setting (defaults to true) rather than inverter_can_charge_during_export - since you've presumably never set the new one (it didn't exist until this PR), Freeze Export is offered exactly as before. It would only stop being offered if you explicitly set inverter_can_freeze_export: false because you'd confirmed your specific inverter can't actually hold SoC flat - not something this change does on its own.