Hello all,
Under certain conditions, I get implausible values from statistics sensors (average linear) after I restart Homeassistant. Implausible means that the values of the statistic sensor suddenly show very high or low temperatures of up to +1000 °C or -1000 °C, although the underlying entity is determined by the outside temperature. My outdoor temperatures tend to range between -10 °C and +42 °C 
.
One of the problematic sensors looks like this:
platform: statistics
name: temperature_forecast_all_next_5d_minimum_average_linear
entity_id: sensor.temperature_forecast_all_next_5d_minimum
state_characteristic: average_linear
precision: 1
sampling_size: 999
max_age:
hours: 168
The temperature_forecast_all_next_5d_minimum_average_linear sensor to which I want to apply the average linear statistic is itself a minimum sensor, which uses values from weather forecast entities.
However, I observe the same problems of extremely high or low temperatures from the statistics sensors (average linear) also when I use actual temperatures from weather integrations as the underlying entity. I tested this for a few weather services and for each sensor there were problems:
- OpenWeatherMap
- Weatherbit
- AccuWeather
- German weather service
- Meteorologisk institutt (Met.no)
My guess is that the statistics sensor (average linear) starts calculating before the internet-based temperature entities have loaded reasonable values. Unfortunately, I do not have the skills to investigate this further.
Does anyone else also observe corresponding problems?
It is perhaps worth mentioning: I also have analogue statistic sensors that are problem-free after restarts. The sensors that work well are based on readings of the temperature that I get via aqara temperature sensors via z2m.