average_power_consumption sensor is mislabeled — value is not a power in W, but appears to be a range correction factor in %

[supernova1963]

The average_power_consumption sensor (battery group, device_class: power, unit W) passes the raw averPowerConsumption field from the electricVehicleStatus cloud response straight through (sign-inverted). After watching this value on my Smart #1 (Premium, 66 kWh) for a full year, I'm fairly confident it is neither a power value nor any kind of consumption:

Evidence:

1. Seasonal behavior is inverted vs. consumption. My car consumes ~17 kWh/100km in summer and up to 24 kWh/100km in winter. The sensor however reads ~75–78 in winter and ~83–85 in summer (see the 12-month long-term statistics chart below). A consumption-like value would move the other way.
2. It matches the range estimation instead. The car's displayed range is strictly linear: range = SoC × (full-charge range / 100). My full-charge range is ~400 km in summer (sensor ≈ 85) and was ~350 km in winter (sensor ≈ 75). Both fit full_range ≈ 470 km nominal × sensor/100 — i.e. the value behaves like the percentage correction factor the BMS applies to a nominal range for the current driving style and temperature conditions.
3. The value only updates while driving, drifts slowly (±5 over days), and stays within ~55–95 over a full year — with the deepest dips on the coldest winter days — consistent with a damped efficiency/calibration factor, not with a measured power (83 W average traction power would be absurd) and not with kWh/100km either.

Suggestion: rename the entity (e.g. "Range efficiency factor"), change the unit to % and drop device_class: power / the W unit, so users don't mistake it for an energy consumption reading. At minimum the current W labeling should go, since it makes HA treat it as a power sensor in dashboards and the Energy view.

Happy to provide my full HA long-term statistics for the sensor if useful.