Problems with trend sensor - works perfectly sometimes, not a all later

tobox · July 23, 2024, 11:40am

To display indicator icons at the temperature values I created trend sensors, that seem to work properly most of the time but the sometimes just “fail”. Any idea what could be the issue? The sensor “schlafzimmer_temperatur_faellt” should definitely be off shortly after 10 o’clock…

template:
  - platform: trend
      schlafzimmer_temperatur_steigt:
        entity_id: sensor.schlafzimmer_thermostat_kanal_2_temperatur
        sample_duration: 600
        max_samples: 5
        min_samples: 2
        min_gradient: 0.00001
  
      schlafzimmer_temperatur_faellt:
        entity_id: sensor.schlafzimmer_thermostat_kanal_2_temperatur
        sample_duration: 600
        max_samples: 5
        min_samples: 2
        min_gradient: -0.00001

Is it possible that there are too few samples coming in during the sample period? But still, that does not really explain the issue for me.

tobox · July 24, 2024, 6:33am

I think I found an explanation:

github.com

home-assistant/core/blob/e1e64be3c9a2597ca17f77ba1ae6c8dbb063eef3/homeassistant/components/trend/binary_sensor.py#L262


      
              self._state = state.state == STATE_ON
          
          async def async_update(self) -> None:
              """Get the latest data and update the states."""
              # Remove outdated samples
              if self._sample_duration > 0:
                  cutoff = utcnow().timestamp() - self._sample_duration
                  while self.samples and self.samples[0][0] < cutoff:
                      self.samples.popleft()
          
              if len(self.samples) < self._min_samples:
                  return
          
              # Calculate gradient of linear trend
              await self.hass.async_add_executor_job(self._calculate_gradient)
          
              # Update state
              self._state = (
                  abs(self._gradient) > abs(self._min_gradient)
                  and math.copysign(self._gradient, self._min_gradient) == self._gradient
              )

If the temperature rises too slow, all steps have a larger distance than “sample_duration”.

Maybe I should switch to the derivative sensor, or change the sample duration.

I think the trend sensor with the current implementation is not very helpful.