Decoding bat_mng_struct.profile... OIDs

[SaxonWood]

In case someone is interesting in the RCT inverters' "forecast", ie. the OIDs:

bat_mng_struct.profile_pdc
bat_mng_struct.profile_pdc_max
bat_mng_struct.profile_load
bat_mng_struct.profile_pext

Note that this is not a real forecast but the inverter's attempt to estimate the future power levels (Pdc, Load, External) based on historic data.

Decoding would require a new non-native type; suggested name BATTERY_MNG_PROFILE = 25.
A possible decoding functions could look like this:

def _decode_bat_mng_profile(data: bytes) -> List[float]:
    '''
    Helper function to decode inverter's internal forecast data series.

    Decodes OIDs named bat_mng_struct.profile_p... 
    Returns an 48-element list of floats, with each float having 
    a power value (Watt). Each item represents a 30 min interval of day, starting at 00:00.
    Data format per 4 octets: float, little endian (DCBA)
    '''
    expected_len = 48*4
    if len(data) != expected_len:
        raise ValueError(f'Forecast payload must be {expected_len} bytes, got {len(data)}')

    fmt = '<48f'                         # format string, repeat 48x
    unpacked = struct.unpack(fmt, data)  # tuple w/ all cell data elements

    return list(unpacked)

More elaborate approaches are of course possible, e.g. returning a dict objects with datetime.time as key and the float as value.

{
    time(0, 0):   0.0,
    time(0, 30): 0.34,
    ...
    time(12, 30): 3456.03,
    ...
    time(23, 30): 0.34,
}

[pluskal]

Confirmed working on a Power Storage 6.0 (BMS SW 5615, 4 battery modules): all four OIDs return 192-byte payloads that decode cleanly with struct.unpack('<48f', data), one Watt value per 30-minute slot starting at 00:00.

Example from a read today — bat_mng_struct.profile_pdc, slots 07:00–10:30: 862, 1443, 1962, 2347, 2834, 3198, 3529, 3640 W; the full day integrates to ~31.1 kWh PV, and profile_load to ~33.9 kWh — both plausible for this site, and the curve start/end matches the local daylight hours.

[mdrmdrmdr]

Does anyone know how the inverter handles the profile array on DST transition days (spring forward / fall back)?

Specifically:

• Does the payload length change (46 or 50 records instead of 48) on the transition day, or does it always return 192 bytes (48 × 4)?
• Or does the record count stay fixed, but the slot-to-time mapping shifts/duplicates a slot around the transition hour?