v0.9.1 - Multiple fixes

Release Notes

---

v0.9.1

Issues: #299, #302, #303, #304, #305, #306, #307, #313

⚠️ BREAKING CHANGE — affects all users. Grid Export Power and Grid Import Power have had their values swapped in all previous versions. After upgrading, each will read the opposite direction to before. Swap any automations, dashboard cards, or Energy Dashboard slots that reference either sensor. Users with Invert Grid Power enabled should also disable it (Settings → Devices & Services → Growatt Modbus → Configure) — it was incorrectly enabled by the setup wizard's auto-detection in previous versions.

New profiles:

• Growatt 3000–15000TL3-S (Issue #299):
  Added support for the TL3-S three-phase grid-tied string inverter (3–15 kW, legacy protocol).
  Register layout confirmed from a full device scan: PV inputs at regs 3–8, total AC power at reg 12,
  per-phase output (R/S/T) at regs 16–25, temperature at reg 32, and energy at regs 53/55 (×0.1 kWh).
  Auto-detected via DTC 2049 at holding register 43. No VPP support — holding register 30000 returns
  Illegal Function. Note: regs 9–10 contain firmware version bytes (not PV2 power); use the
  Total PV Power sensor rather than PV2 Power.

• MIC 2500–5500MTL-S (Issue #304):
  Added support for the MIC 2500–5500MTL-S single-phase dual-string grid-tied inverter family
  (2.5–5.5 kW, legacy V3.05 protocol). Inherits the MIC 600–3300TL-X register map with a second
  PV string confirmed at regs 7–8. Auto-detected via DTC 210. Note: this inverter rejects any
  Modbus read of more than one register at a time — the max_block_size: 1 mode introduced in
  this release handles this automatically for all sensor reads.

• MID Hybrid (11–30kW) (Issue #313 / PR #314):
  MID 11–30KTL3-XH and MID 8–15KTL3-XHL/JP share DTC 5400 with MOD 3–10KTL3-XH and use identical
  register layouts. A named manual-selection option ("MID Hybrid (11–30kW)") is now available in the
  profile dropdown for MID users who prefer the MID label. Auto-detection continues to route DTC 5400
  to the MOD Hybrid profile, preserving entity IDs for existing users.

Fixes:

• Breaking fix: grid_export_power and grid_import_power swapped on all profiles (Issue #302):
  Both always-positive derived power sensors had inverted formulas in all previous versions.
  On hybrid profiles (SPH, MOD, WIT) the symptom was visible: during grid import,
  grid_export_power showed the import magnitude while grid_import_power read zero.
  On grid-tied string inverters (MIN, MIC, MID), grid_export_power silently read zero and
  grid_import_power carried the export value under the wrong name.
  The signed grid_power sensor and all daily energy sensors were unaffected.
  Also fixes the setup wizard's grid orientation detection, which was enabling Invert Grid Power
  for the wrong case — inverting when it should not, and not inverting when it should.

• Fix: MIC 2500–5500MTL-S entities all unavailable (Issue #304):
  The inverter rejects any Modbus read of more than one register at a time
  (ExceptionResponse FC=132, exception_code=1 — Illegal Function for any block read).
  Fixed by adding per-profile max_block_size support to the coordinator. When a profile
  sets max_block_size: 1, the coordinator switches to sparse read mode: only the specific
  register addresses defined in the profile are read, grouped into consecutive runs of at most
  1 register, skipping all gaps. This eliminates unnecessary reads for undefined registers.
  All other profiles continue to use the original dense 125-register block reads unchanged.

• Fix: Inverter Status shows wrong text on hybrid inverters (Issue #305):
  The status sensor was using a single status code table shared across all inverter families.
  Hybrid inverters (SPH, SPM, MOD, WIT, MIN TL-XH, SPA, SPE) use the VPP Protocol V2.01
  status map where codes have completely different meanings to the grid-tied map — most critically,
  code 5 was shown as "Standby" when the correct label is "PV On-Grid", and code 1 was shown as
  "Normal" when it means "Self-Test" on hybrid models. SPF off-grid inverters have their own
  distinct status set which was also mixed into the same dict.
  Fixed by introducing three separate status code tables (STATUS_CODES for grid-tied,
  HYBRID_STATUS_CODES for hybrid, SPF_STATUS_CODES for off-grid) and selecting the correct
  table at runtime based on the active profile.

  Code	Grid-tied	Hybrid	SPF Off-Grid
  0	Waiting	Waiting	Standby
  1	Normal	Self-Test	—
  2	—	Reserved	Discharge
  3	Fault	Fault	Fault
  4	—	Updating	Bypass
  5	—	PV On-Grid	PV Charge
  6	—	Bat On-Grid	Combined Charge
  7	—	PV+Bat Off-Grid	—
  8	—	Bat Off-Grid	—
  9	—	Bypass	—

• Fix: SPH/SPM 8000–10000TL-HU auto-detection fails (Issue #303):
  DTC code 21303, which identifies the SPH/SPM 8000–10000TL-HU running firmware UL2.21,
  was not in the DTC map. Detection fell through to the legacy register-probing path, which could
  misidentify the high-PV-voltage SPH HU as a MIC micro inverter. On the MIC profile, all battery
  and grid registers are absent, so every power-flow derived sensor was computed from solar
  generation alone. Fixed by mapping DTC 21303 → sph_8000_10000_hu.

• Fix: house_consumption returns solar generation on SPH/SPM HU (Issue #303):
  Two related issues caused house_consumption to equal solar on HU firmware UL2.21.
  First, the HU variant does not populate power_to_load (register 1021/1022).
  Second, the fallback attempted self_consumption_power (register 1037/1038), which on UL2.21
  reports solar-only self-consumption rather than total house load.
  Fixed by removing the intermediate step: when power_to_load is absent the integration now
  goes directly to the full energy balance
  (solar + battery_discharge − battery_charge + grid_import − grid_export)
  using the direct grid registers (1015/1016 and 1029/1030), which are correct on UL2.21.

• Fix: Remote Charge/Discharge Power slider rejects negative values (Issue #306):
  Register 30409 (remote_charge_and_discharge_power) accepts −100% (full discharge) to +100%
  (full charge). Negative values must be sent as unsigned 16-bit two's complement over Modbus
  (e.g. −80 → 65456 / 0xFFB0). The code was passing the raw negative Python integer directly to
  pymodbus, which expects an unsigned 16-bit value and raises a struct error. Positive values
  worked because no conversion was needed.
  Fixed by converting signed values to unsigned 16-bit (value & 0xFFFF) in write_register
  before the pymodbus call. The read-back verification comparison in write_register_verified
  is also updated to compare against the unsigned representation.

• Fix: energy_today rises through the night on SPH hybrid inverters (Issue #307):
  The hardware register for energy_today (reg 53/54) on SPH/WIT hybrid inverters counts total
  AC system output, including battery discharge — not solar generation only. The integration
  normally avoids this by summing the per-MPPT DC string registers (pv1_energy_today + pv2_energy_today), which track PV input only. However, the guard condition checked
  pv*_energy_today > 0. After the inverter's daily counters reset to zero at midnight, all MPPT
  values return 0, the guard evaluated False, and the code fell back to register 53/54 — which
  then climbed through the night at a rate matching house load (battery powering the house).
  Fixed by gating on register address existence rather than value > 0. If the profile defines
  per-MPPT energy registers (which all SPH/WIT profiles do), the MPPT sum is always used —
  including when it is zero. Zero at night is the correct value.

• Fix: MID 15–25kW PV3 missing from base profile (Issue #313):
  The base MID_15000_25000TL3_X profile was missing PV3 registers 11–14. PV3 was previously
  only available on the V2.01 variant via VPP registers 31018–31021. Base register availability
  confirmed from the Issue #313 scan. MID models with a third MPPT string will now show
  pv3_voltage, pv3_current, and pv3_power on both base and V2.01 profiles.

• Docs: corrected legacy storage register 1000 (uwSysWorkMode) status description:
  Previously showed 0x05-0x08=Normal. Updated to list each code individually, matching the
  VPP Protocol V2.01 register 31000 definition.

---

v0.9.1b6

Issues: #307

• Fix: energy_today rises through the night on SPH hybrid inverters (Issue #307):
  On SPH (and WIT) hybrid inverters the hardware register for energy_today (reg 53/54)
  counts total AC system output, including battery discharge — not solar generation only.
  The integration normally avoids this by summing the per-MPPT DC string registers
  (pv1_energy_today + pv2_energy_today), which track PV input only.

  However, the guard condition checked pv*_energy_today > 0 before using the MPPT sum.
  After the inverter's daily energy counters reset to zero at midnight, all MPPT values
  return 0, the guard evaluated False, and the code fell back to register 53/54 — which
  then climbed through the night at a rate matching house load (battery powering the house).

  Fixed by gating on register address existence rather than value > 0. If the profile
  defines per-MPPT energy registers (which all SPH/WIT profiles do), the MPPT sum is always
  used — including when it is zero. Zero at night is the correct value.

---

v0.9.1b5

Issues: #303, #306

• Fix: Remote Charge/Discharge Power cannot be set to negative values (Issue #306):
  Register 30409 (remote_charge_and_discharge_power) accepts values from −100% (full discharge)
  to +100% (full charge). Writing a negative value requires sending the unsigned 16-bit
  two's complement representation over Modbus (e.g. −80 → 65456 / 0xFFB0). The code was
  passing the raw negative Python integer directly to pymodbus, which expects an unsigned
  16-bit value and raises a struct error. Positive values worked because no conversion was
  needed.

  Fixed by converting signed values to unsigned 16-bit (value & 0xFFFF) in write_register
  before the pymodbus call. The read-back verification comparison in write_register_verified
  is also updated to compare against the unsigned representation, so positive write + verified
  confirmation is correctly reported.

• Fix: SPH/SPM 8000-10000TL-HU auto-detection fails (DTC 21303 unmapped, Issue #303):
  DTC code 21303, which identifies the SPH/SPM 8000–10000TL-HU running firmware UL2.21,
  was not in the DTC map. The detection logic could fall through to the legacy register
  probing path, which may identify a high-PV-voltage SPH as a MIC micro inverter. On the
  MIC profile, all battery and grid registers are absent, so every power-flow derived sensor
  (Grid Export Power, Grid Import Power, House Consumption) was computed from solar
  generation alone.

  Fixed by adding 21303 → sph_8000_10000_hu to the DTC map. Detection now resolves at
  Step 1 (OffGrid DTC) and reaches the correct profile without reaching the register-probing
  fallback.

• Fix: house_consumption returns solar on SPH HU regardless of energy balance (Issue #303):
  Register 1037/1038 (self_consumption_power) on HU firmware UL2.21 reports solar-only
  self-consumption, not total house load. The previous code used it as a direct house load
  proxy when power_to_load (register 1021/1022) was zero, causing house_consumption to
  equal solar in most operating conditions. The energy balance
  (solar + discharge − charge + import − export) was never reached.

  Fixed by removing the self_consumption_power intermediate step. When power_to_load = 0,
  the integration now goes directly to the energy balance using the direct grid import/export
  registers (1015/1016 and 1029/1030), which are correct on UL2.21.

---

v0.9.1b4

Issues: #305

• Fix: Inverter Status shows wrong text on hybrid inverters (Issue #305):
  The Inverter Status sensor was using a single status code table shared across all inverter
  families. Hybrid inverters (SPH, SPM, MOD, WIT, MIN TL-XH, SPA, SPE) use the V1.39 / VPP
  Protocol V2.01 status map where the codes have completely different meanings to the grid-tied
  map — most critically, code 5 was shown as "Standby" when the correct label is "PV On-Grid",
  and code 1 was shown as "Normal" when it means "Self-Test" on hybrid models. SPF off-grid
  inverters have their own distinct status set (0=Standby, 2=Discharge, 5=PV Charge, etc.)
  which was previously mixed into the same dict and would now also collide.

  Fixed by introducing three separate status code tables (STATUS_CODES for grid-tied,
  HYBRID_STATUS_CODES for hybrid, SPF_STATUS_CODES for off-grid) and selecting the correct
  table at runtime based on the active profile. Hybrid status codes are now:

  Code	Label
  0	Waiting
  1	Self-Test
  2	Reserved
  3	Fault
  4	Updating
  5	PV On-Grid
  6	Bat On-Grid
  7	PV+Bat Off-Grid
  8	Bat Off-Grid
  9	Bypass

• Docs: corrected legacy storage register 1000 (uwSysWorkMode) status description:
  The protocol-v139 reference table previously showed 0x05-0x08=Normal for register 1000.
  Updated to list each code individually, matching the VPP Protocol V2.01 register 31000
  definition which was already correct in the docs.

---

v0.9.1b3

Issues: #304

• Fix: MIC 2500-5500MTL-S entities all unavailable (Issue #304):
  The inverter rejects Modbus reads of more than one register at a time (responds with
  ExceptionResponse FC=132, exception_code=1 — Illegal Function for any block read).
  Fixed by adding per-profile max_block_size support to the coordinator. When a profile
  sets max_block_size: 1, the coordinator switches to sparse read mode: it reads only the
  specific register addresses defined in the profile, grouped into consecutive runs of at most
  max_block_size registers, skipping all gaps. This also eliminates unnecessary register reads
  for registers not defined in the profile. All other profiles continue to use the original
  dense read mode (contiguous 125-register block reads) unchanged.

---

v0.9.1b2

Issues: #299, #303, #304

• New profile: Growatt 3000-15000TL3-S (Issue #299):
  Added support for the TL3-S three-phase grid-tied string inverter (3–15 kW, legacy protocol).
  Register layout confirmed from a full device scan: PV inputs at regs 3–8 (PV1 voltage/current/power
  and PV2 voltage/current), total AC power at reg 12, per-phase output (R/S/T) at regs 16–25,
  temperature at reg 32, and energy at standalone regs 53/55 (×0.1 kWh). Auto-detected via DTC 2049
  at holding register 43. No VPP support — holding register 30000 returns Illegal Function.
  Note: regs 9–10 contain firmware version bytes (not PV2 power), so PV2 power stays 0; use the
  Total PV Power sensor instead.

• Fix: house_consumption mirrors solar on SPM 8000TL-HU (Issue #303):
  The HU variant of the SPH family does not populate power_to_load or power_to_user registers,
  causing house_consumption to collapse to solar power when both are zero (the simple load = solar
  fallback). Fixed with a full energy balance fallback: when battery or grid data is present,
  house_consumption = solar + battery_discharge − battery_charge + grid_import − grid_export.
  With the HU test case (solar 2389 W, import 2801 W, discharge 501 W) this now correctly
  computes ≈ 5691 W instead of 2389 W.

• New profile: MIC 2500-5500MTL-S (Issue #304):
  Added support for the MIC 2500–5500MTL-S single-phase dual-string grid-tied inverter family
  (2.5–5.5 kW, legacy V3.05 protocol). Inherits the MIC 600–3300TL-X register map with PV2
  voltage and current confirmed at regs 7–8. PV2 power pair at regs 9–10 is included but
  unconfirmed — if the PV2 Power sensor shows an unrealistic value, those registers contain
  firmware version bytes and should be reported via the issue tracker.
  Auto-detected via DTC 210 at holding register 43. Note: this inverter rejects multi-register
  Modbus block reads — use block_size=1 with the Universal Scanner service.

---

v0.9.1b1

Issues: #302

We apologise for this one. The Grid Export Power and Grid Import Power sensors have had
their values swapped since they were introduced, meaning every user has been seeing export
labelled as import and vice versa. The bug was in the formula that splits the signed grid power
value into its two always-positive halves — the positive and negative parts were extracted into
the wrong sensors. The signed Grid Power sensor and all daily energy sensors were unaffected
because they use a different code path. We only caught it because a user noticed the symptom
on a hybrid inverter where both directions are active (Issue #302). Grid-tied string inverter
users (MIN, MID, MIC) were also affected but the symptom was less obvious — those inverters
only export during the day, so Grid Export Power silently read zero while Grid Import Power
carried the export value. Both are now correct.

⚠️ BREAKING CHANGE — affects all users.
Grid Export Power and Grid Import Power were swapped in all previous versions.
After upgrading, these sensors will read the opposite value to before.
If you have automations, dashboards, or energy dashboard slots referencing either sensor,
you must swap them. The signed Grid Power sensor and all daily energy sensors are unaffected.
Users with Invert Grid Power enabled should also disable it (Settings → Devices &
Services → Growatt Modbus → Configure) — it was incorrectly set by auto-detection in
previous versions and is no longer needed for standard inverters.

• Breaking fix: grid_export_power and grid_import_power swapped on all profiles (Issue #302):
  The two always-positive derived power sensors had their formulas inverted in all previous
  versions. grid_export_power was extracting the import portion of the signed grid power
  value and grid_import_power was extracting the export portion — the opposite of their names.
  On hybrid profiles (SPH, MOD, WIT) the symptom was visible: during grid import,
  grid_export_power showed the import magnitude while grid_import_power read zero. On
  grid-tied string inverters (MIN, MIC, MID) the inverter only exports, so grid_export_power
  silently read zero and grid_import_power showed the export value under the wrong name.
  The signed grid_power sensor and the daily energy sensors (Energy to Grid Today,
  Grid Import Energy Today) were unaffected — they read from separate register addresses.

• Fix: invert_grid_power auto-detection was inverting the wrong case:
  The setup wizard's grid orientation detection returned the wrong result — it enabled inversion
  when it detected positive = export (the correct convention) and disabled it when it detected
  negative = export (the case that actually needs correction). This caused some users to have
  invert_grid_power incorrectly set to True, which negated the signed grid_power sensor
  so it showed negative while exporting. Auto-detection is now correct. Existing users with
  the flag incorrectly enabled should disable it via Configure.

---