Hi,
I’ve got an Hichi IR sensor at my power meter. This delivers bad data every now and then. Sometimes it knows it’s wrong and delivers 999999. sometimes it doesn’t realize a single digit was flipped.
Step by step I eliminated the errors by adding some IFs:
state: >
{% set new = states('sensor.tasmota_emh_e_in') | float(999999) %}
{% set old = this.state | float(0) %}
{% if new == 999999 or new > 10000 %}
{{ old }}
{% else %}
{% if new < old %}
{{ old }}
{% else %}
{% if new > old + 2000 %}
{{ old }}
{% else %}
{{ new }}
{% endif %}
{% endif %}
{% endif %}
Now I deleted the bad measurements from the STATE table using the SQLLite app. But they do not disappear from graphs which makes graphs unreadable when there is a 2MW spike or something like that.
I tried fiddling with the statistics table but could not find any wrong data.
any idea how to eliminate bad data from graphs?
thanks
PS: shouldn’t total_increasing sensors drop data that’s below the last data point?