[1.5.0b5] - 2026-09-30
Added
- Remove any battery from the options flow (Batteries → Remove a battery): pick it by name; its device and entities are deleted too. Before, only the last battery could be dropped, and deleting the device was rejected.
Fixed
- Anker E5000 with a CT meter no longer caps the fleet discharge: its AC contribution is now pack minus PV, not register 10012 (grid reading) (#468).
- Dynamic Pricing no longer re-buys the same deficit across closely following slots: it charged to the ceiling instead of the planned energy (#518).
- Overcast days no longer reset a mature solar profile: capacity-change detection now compares each day's production with its forecast instead of raw peaks, so grey days the forecast predicted are not taken as lost capacity.
- A BMS that cuts off below 100% while flapping Charge ↔ Standby is now recognised as full: a single accepted sample reset the cutoff detector, so a Venus E v3 stuck at 88% with every cell full was commanded to charge forever.
- Sunrise and solar noon now match the almanac (#410, #516): computed with
astral(as HA'ssun.sun) instead of an approximate formula that ran up to ~27 min late. Thanks to @syphernl. - A normal top-of-charge taper is no longer treated as a dead battery: the last stretch before 100% can drop to a fraction of the commanded power well before the BMS confirms a cutoff (Standby + ≤10 W), which the non-delivery tracker read as a fault and answered with repeated wake nudges (RS485 re-assert) even though SOC kept climbing. A battery charging at SOC ≥ 99% now gets up to 45 minutes to finish tapering before being judged non-responsive; one genuinely stuck in that band still surfaces as a fault, just later (#525). Thanks to @sphings79.
- A full battery (100% SOC or confirmed BMS cutoff) no longer keeps a stale charge-side "battery not delivering" Repair open; discharge and comms faults are left alone (#524). Thanks to @sphings79.