Can someone please suggest ideas for how to reliably detect these high plateaus in this sensor i have?
They are not always at the same level. So far i have average_step statistic and then check if it exceeds 200k value but that does not work reliably. Is there some sort of better way to do it? They always start with big discontinuos jump in value so maybe some sort of spectral filter - need help….
If you can pick a threshold, it’s then easy to create a binary sensor when the signal exceeds the threshold for more than x minutes.
Alternatively create a template sensor that triggers off the signal and calculates the difference between the trigger to value and the trigger from value. This is the amount of change. You can then create a binary sensor that goes to True when this exceeds a threshold and stays True until it exceeds a negative threshold.
Or you can combine both together.
What is that sensor measuring?