Off-by-one minute error in work mode slot time calculations causing invalid Growatt TL3-XH time slot values

[tomasvonhofsten]

What happened

BESS Manager consistently sends work mode slot end times that are 1 minute early when controlling Growatt TL3-XH inverters via the Growatt Server HA integration.

Examples observed:

• Slot 1 end: 20:44 (should be 20:45)
• Slot 2 end: 12:59 (should be 13:00)
• Slot 3 end: 16:14 (should be 16:15)
• Slot 4 end: 17:14 (should be 17:15)

The Growatt integration only accepts 15-minute increment values
(00:00, 00:15, 00:30... etc). When BESS sends an off-by-one value
like 16:14, HA rejects it with:

"Invalid option for select.growatt_tl_x_xh_tl3_xh_work_mode_slot_3_end:
'16:14'"

This means the intended schedule is never actually applied to the
inverter — the work mode slots remain at their previous values.
BESS Manager believes it has set the schedule but the inverter is
not following it.

Setup:

• Inverter: Growatt MOD 10KTL3-XH
• Battery: Growatt ARK 15kWh
• HA Growatt Server integration with API token auth
• BESS Manager 8.6.0

Debug bundle

Please drag-and-drop bess-debug-2026-05-19-080021.md from your Downloads folder into the editor below before submitting.
The file contains your version, settings, recent logs, and the raw HA discovery dump that the maintainers need to reproduce the issue.

bess-debug-2026-05-19-080021.md

[johanzander]

@claude-bot analyze

[bess-manager-claude-bot]

Root cause

GrowattScheduleManager._groups_to_tou_intervals() computes TOU slot end times using an inclusive last-minute formula — it takes the start of the final period in a group and adds 14 minutes to get the last minute of that period (e.g. period starting at 16:00 → end_time = 16:14). The Growatt TL3-XH exposed via the Growatt Server HA integration requires end times to be on exclusive 15-minute boundaries (the start of the next period) — only HH:00, HH:15, HH:30, and HH:45 are valid values. The +14 formula always produces an answer exactly 1 minute before the required boundary, so every end time is rejected by HA with an "Invalid option" error, silently leaving the inverter at its previous schedule.

Evidence

• core/bess/min_schedule.py:480–482 — hardware-critical path where the bad end time is computed:

end_hour, end_minute = self._period_to_time(group["end_period"])
# End minute should be the last minute of the period (14, 29, 44, or 59)
end_minute = end_minute + 14

• core/bess/min_schedule.py:511 — where the bad value is formatted into the interval sent to the inverter:

"end_time": f"{end_hour:02d}:{end_minute:02d}",

• core/bess/min_schedule.py:413 — same +14 pattern in the display/detail path (not hardware, but shows the same assumption):

end_m += 14  # Last minute of period

• core/bess/battery_system_manager.py:170–172 — confirms GrowattScheduleManager (min_schedule.py) is the code path used for non-SPH inverters including the TL3-XH:

if _inverter_type == "SPH":
    return SphScheduleManager(battery_settings=self.battery_settings)
return GrowattScheduleManager(battery_settings=self.battery_settings)

The pattern in the report is perfectly consistent with the formula: every wrong time is exactly 1 minute before a 15-minute boundary (16:14 instead of 16:15, 12:59 instead of 13:00, 20:44 instead of 20:45, 17:14 instead of 17:15).

Proposed fix

In _groups_to_tou_intervals() (lines 480–482 of min_schedule.py), replace the +14 inclusive calculation with _period_to_time(group["end_period"] + 1). This returns the start of the period immediately after the last group period, which is the exclusive 15-minute boundary the TL3-XH expects. The existing DST overflow guard (if end_hour >= 24: cap to 23:59) already handles the midnight edge case correctly and requires no change. The same substitution should be applied to the display path at line 413 and the logging path at line 846 for consistency, though only the hardware path at line 482 causes the actual HA rejection.

Risks / open questions

• MIN-series inverters: The original MIN inverter models (the class is named GrowattScheduleManager after the MIN series) may have used the older custom integration that accepted inclusive HH:14/HH:29/HH:44/HH:59 values. The fix changes the emitted format to exclusive boundary values — a regression test against MIN-series behavior would be prudent, or confirmation that all MIN users also run Growatt Server integration.
• Last period edge case: If the last group ends at period 95 (23:45–23:59), _period_to_time(96) returns (24, 0), which the existing end_hour >= 24 cap will correctly clamp to 23:59. This is intentional and unchanged.
• Display path consistency: The +14 in get_detailed_period_groups() (line 413) is display-only and doesn't affect hardware, but after the fix the displayed end times will differ from the hardware times if left unchanged. Updating both paths is recommended.

---

Reply @claude-bot fix to implement and open a draft PR.

[johanzander]

Thanks for the detailed report and debug bundle, @tomasvonhofsten. I took a closer look and I think there are a few things going on here that the automated analysis missed.

The +14 end times are intentional

BESS Manager uses inclusive end times for TOU intervals — 16:00–16:14 means "this slot covers the 15 minutes from 16:00 through 16:14." The next slot starts at 16:15. This avoids overlapping intervals, which is a hard requirement of the Growatt inverter hardware. The design is documented at the top of min_schedule.py:

"Uses hour boundaries (20:00-20:59, 21:00-21:59) — no overlap possible"

This is working as designed on all current installations.

The error you are seeing is not from BESS Manager

The error message you reported references a select entity:

select.growatt_tl_x_xh_tl3_xh_work_mode_slot_3_end

BESS Manager does not use select entities for TOU control. It calls the growatt_server.update_time_segment service, which accepts arbitrary HH:MM values. The "Invalid option" error is coming from something else interacting with those select entities — possibly an automation, a dashboard card, or the Growatt Server integration exposing the TL3-XH slots as select entities with a fixed option list.

Your debug bundle shows a more fundamental issue

The debug bundle reveals that most battery-related sensors are reporting "unavailable" — the Growatt device is not communicating properly with Home Assistant. The system also cannot read InfluxDB historical data and reports "No hourly settings for hour 0." With sensors unavailable, BESS Manager cannot read or control the inverter regardless of the time format.

Suggested next steps

1. Fix sensor availability first. Check that the Growatt Server integration is online and the device (QDL3DA90YP) is reachable. Until sensors report values, nothing else will work.
2. Check what is writing to those select entities. BESS Manager uses the growatt_server.update_time_segment service call, not select.select_option. If something else is trying to set those select entities with the TOU times, that is the source of the "Invalid option" errors.
3. Once sensors are back, BESS Manager should be able to apply schedules normally via the service call path.

I am closing the automated analysis on this one since the root cause differs from what was posted. Let me know how it goes with the sensor connectivity — happy to help debug further.

[johanzander]

Closing this — the bot's analysis was incorrect on multiple points:

1. The +14 inclusive end times are intentional design, not a bug. Growatt TOU slots use inclusive minutes to avoid overlapping intervals between adjacent periods.
2. The "Invalid option" error is irrelevant — it refers to select entities, but BESS uses the growatt_server.update_time_segment service, not select entities.
3. The actual symptom (sensors showing "unavailable") is a Growatt cloud/integration issue, not something caused by BESS.

No code change needed.