Discharge level

[Doodlehusse]

bess-debug-2026-06-29-170818.md

What happened

Ver 9.8.1. Discharge level does not change according to the inverter plan when work mode is Load First. Changes correctly when work mode is Grid first.

Debug bundle

Please drag-and-drop bess-debug-2026-06-29-170818.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.

[Doodlehusse]

Discharge level change correctly according to the inverter plan in work mode Load First when using Growatt upstream integration but not when using Solax Modbus!

[johanzander]

Will have a look tomorrow when my weekly Claude credits fill up again..

[bess-agent]

Root cause

On the SolaX Modbus / local-Modbus integration path (SolaxModbusGrowattController), the planned discharge rate for LOAD_SUPPORT periods is never written to the inverter. apply_period() skips the entire hardware write (both grid_charge and discharge_rate) whenever the derived TOU mode is load_first. That gate was added deliberately in #166 to stop a real bug — writing discharge_rate=0 during IDLE/SOLAR_STORAGE was disabling the inverter's native self-use discharge — but it gates on mode, not intent. Four strategic intents map to load_first (IDLE, SOLAR_STORAGE, SOLAR_EXPORT, LOAD_SUPPORT); the first three correctly want discharge_rate=0, but LOAD_SUPPORT computes a real, non-zero action-derived discharge percentage that gets silently dropped by the same gate. BATTERY_EXPORT maps to grid_first, bypasses the gate, and writes correctly — which is exactly why Grid First works and Load First doesn't.

Evidence

• core/bess/solax_modbus_growatt_controller.py:151-159:

  # load_first uses inverter-native discharge control; touching the EMS
  # discharge rate register (even with 0) disables discharge entirely.
  if mode != "load_first":
      success, error_msg = self._write_period_to_hardware(
          controller, grid_charge, discharge_rate
      )

• core/bess/inverter_controller.py:52-59 — INTENT_TO_MODE maps LOAD_SUPPORT to load_first, same as the zero-rate intents:

  INTENT_TO_MODE: ClassVar[dict[str, str]] = {
      "GRID_CHARGING": "battery_first",
      "SOLAR_STORAGE": "load_first",
      "LOAD_SUPPORT": "load_first",
      "BATTERY_EXPORT": "grid_first",
      "SOLAR_EXPORT": "load_first",
      "IDLE": "load_first",
  }

• core/bess/inverter_controller.py (_map_intent_to_rates) — LOAD_SUPPORT computes a non-zero discharge_rate from planned battery power when discharging, i.e. this is a real write that matters, not a no-op:

  elif intent == "LOAD_SUPPORT":
      if battery_action_kw < -0.01:
          discharge_rate = min(100, max(0, round(abs(battery_action_kw) / self.max_discharge_power_kw * 100)))
      else:
          discharge_rate = 0
      return False, discharge_rate

• Reporter's debug bundle (v9.8.1, inverter_type = "solax_modbus_growatt_min"): at export time number.solaxa_inverter_ems_discharging_rate = 0 while sensor.solaxa_inverter_battery_discharge_power = 316.0 W (actively discharging) — the register is stale/unmanaged even though the battery is discharging, consistent with the write being skipped.
• core/bess/tests/unit/test_solax_modbus_growatt_single_segment.py:208-222 — the regression test added with fix: skip EMS discharge rate write in load_first mode (SolaxModbus) #166 only covers IDLE (where discharge_rate=0 is correct); there is no test covering LOAD_SUPPORT's non-zero rate under load_first, which is why this gap wasn't caught.

Proposed fix

Narrow the gate in apply_period() so it distinguishes "no discharge rate needs writing" from "mode is load_first" — e.g. only skip the hardware write when discharge_rate == 0 (or more precisely, skip only the write that would zero out the EMS register), rather than skipping unconditionally for any load_first-mapped intent. LOAD_SUPPORT's non-zero, action-derived rate should still reach the inverter. Add a regression test covering LOAD_SUPPORT under load_first mode to close the gap left by #166's test coverage.

Risks / open questions

• Not verified from the debug bundle: what the inverter does physically with a stale EMS register while in load_first — whether native self-use logic overrides it entirely or partially respects the last-written value. The bundle only proves the app never updates the register during LOAD_SUPPORT, not the downstream physical effect.
• The debug bundle's log window (00:00–~04:15) only covers load_first-mapped intents, so there's no in-bundle before/after comparison of a load_first write skip vs. a grid_first write; the mode/intent asymmetry is established from code, not from a direct log contrast.
• Fix should preserve fix: skip EMS discharge rate write in load_first mode (SolaxModbus) #166's original protection (never zero out the EMS register during IDLE/SOLAR_STORAGE/SOLAR_EXPORT) while unblocking LOAD_SUPPORT's legitimate non-zero writes.

---

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