Open seniskenis opened 1 year ago
And totally wrong behavior on this set of data: duration: 2 moving: search_length: 24
Thanks for sharing, I have not been able to test this very much since it requires you to debug during specific hours of the day unless running it in a simulated environment which I have not found time to create. As for my own use I'm not using this very much at the moment as boiler-room is under renovation and no wifi access to heat-pump at the moment, hence I have not been checking the result very much lately.
For the but I could imagine it could miss to check current_hour, hence duration=1 would give only one possible slot. Could also be related to #12 where it's off-by-one.
Cannot make any promises to investigate this now as I'm busy with the renovation but will keep in mind as soon as done.
Please check if release 0.2.3 fixes the issue with duration = 1
This appears to still be an issue even today...
From the last plot above I would say the following: 00:00-01:00 On -> Correct as the average price 00:00-02:00 likely the lowest 01:00-05:00 Off -> Correct as the average 05:00-07:00 lowest ahead 05:00-08:00 On -> Correct as price is continuously increasing and hence no cheaper average can be found in the future (the moving window does not look in the past) 08:00-09:00 On -> Questionable as visually I would say the average 22:00-24:00 look to be lower
In documentation: "duration can be in the range of 1 to 5 and specifies how large window of consecutive hours to slide forward in search for a minimum average price."
in my case I don't need hours to be consecutive, it can be 02:00, 04:00, 05:00... etc. so "duration = 1" would be good for me, but when is set to one, the sensor is always ON.
in case: and duration set to 2 sensor starts at 02:00, but the good price is also at 00:00 so one hour remains unused.
can "duration = 1" bug be fixed or I'm doing something wrong, because it always on for me.
edit: after midnight strategy changed and with "duration=2" all consecutive hours from 00:00 to 07:00 were used, so this time it worked ok. but another problem happened in post bellow.