Filter unwanted values / spikes from data

[Prutsium]

Hi all,

I am running here on my heater system an additional ESP to calculate the Delta-T from the water in / out so that i would be able to keep this at an optimal level so condensation (gas heater) is optimal.
This all works great by reading the in / out temperature and substracting them and dump the result into a dummy sensor.

But from time to time i get some nasty spikes and since all is stored into a DB to create visualisation with Grafana it looks not so nice

My idea is to filter these spikes out somehow with a potential rule within the ESP but i could not find anything specific then just do an action if value = > XX Level

The setup:
ESP: (See the black marked area's)


Grafana: (See the red marked area these are the spikes i want to get rid of)


The current rule to fill the Dummy sensor:
On Living_Radiator_In#Temp_3 do
TaskValueSet 7,1,[Living_Radiator_In#Temp_3]-[Living_Radiator_Out#Temp_4]
endon

Question is: How to filter these sudden drops ....

Thanks!
Jerome

[grovkillen]

If it's always a one point that is extremely off then you can use one of the left over dummy values. Add the data point to that one and compare the new data point with that dummy. If the value differs more than a set threshold you just skip to publish it to the database.