Hi,
I’m planning to use Grafana and influx DB to visualise statistics for my heat pump over the course of a year. I found a pretty good instruction that I can certainly execute. One aspect that is recommended is to really review which data needs to be captured. This is a good approach in general, and so I took the opportunity.
But in this context, I discovered something weird:
I have a sensor that captures the COP for the heat pump once a day. This works as expected every night at 23:55. Configuration:
- trigger:
- platform: time
at: "23:55:00"
sensor:
- name: "COP WW täglich"
state: >
{% set heat = states('sensor.erzeugte_warmemenge_ww') | float(0) %}
{% set energy = states('sensor.energie_verbrauch_ww') | float(0) %}
{{ (heat / energy) | round(2, default=0) }}
availability: >
{{states('sensor.erzeugte_warmemenge_ww')|is_number and states('sensor.energie_verbrauch_ww')|is_number}}
state_class: measurement
However, when looking at the history for that sensor it has actually more data points than originally captured. This is because the long term statistics table is inflated by hourly average values which are pointless for such a sensor. Check the data in this table:
| Sensor | Value | Date |
|---|---|---|
| sensor.cop_ww_taglich | 4,7799844145 | 2024-09-25T21:00:00.000Z |
| sensor.cop_ww_taglich | 4,89 | 2024-09-25T22:00:00.000Z |
| sensor.cop_ww_taglich | 4,89 | 2024-09-25T23:00:00.000Z |
| sensor.cop_ww_taglich | 4,89416018554167 | 2024-09-26T21:00:00.000Z |
| sensor.cop_ww_taglich | 4,94 | 2024-09-26T22:00:00.000Z |
| sensor.cop_ww_taglich | 4,94 | 2024-09-26T23:00:00.000Z |
| sensor.cop_ww_taglich | 4,94582424200556 | 2024-09-27T21:00:00.000Z |
| sensor.cop_ww_taglich | 5,01 | 2024-09-27T21:55:00.467Z |
| sensor.cop_ww_taglich | 4,75 | 2024-09-28T21:55:00.466Z |
| sensor.cop_ww_taglich | 4,94 | 2024-09-29T21:55:00.467Z |
| sensor.cop_ww_taglich | 4,71 | 2024-09-30T21:55:00.466Z |
| sensor.cop_ww_taglich | 5,15 | 2024-10-01T21:55:00.467Z |
| sensor.cop_ww_taglich | 4,6 | 2024-10-02T21:55:00.466Z |
| sensor.cop_ww_taglich | 5,01 | 2024-10-03T21:55:00.466Z |
| sensor.cop_ww_taglich | 4,86 | 2024-10-04T21:55:00.466Z |
| sensor.cop_ww_taglich | 4,82 | 2024-10-05T21:55:00.467Z |
So instead of reducing the stored data, the transformation of the long term statistics inflates the data.
Has anyone an idea how to circumvent this?
Some ideas from my end:
- Ignore it since Influx DB does anyway not depend on the long term statistics
- Change the state_class of the sensor from measurement to ???