Unclear on fixing false spikes with lifetime total energy

[still scaffold]

I have huge spike artifacts in my energy generation panel. As far as I know, this happens because it draws on "lifetime total energy", which is supposed to only ever increase. The increase between two energy measurements is then read as the energy produced in the time period between the two measurements. But sometimes, the entity will (for whatever reason) erroneously, momentarily report 0, which is interpreted as a reset on the counter and then when the next value continues as normal, that is interpreted like all of its power ever was generated since that 0 "measurement".

I also know that I am supposed to be able to fix this using the statistics panel, clicking the "adjust" icon and having it find all outliers for me to fix.

But what I don't understand is, why are all these outliers the high values, when I am editing "lifetime energy"? Shouldn't the outliers be the 0s? In general, why do the values on the "lifetime energy" appear to NOT be lifetime energy at all?

[craggy sapphire]

Statistics store the change which is what you are supposed to correct. The absolute value doesn't really matter for any of the graphs but it is stored and recomputed alongside

[still scaffold]

It's unfortunate because the absolute value is easier to correct, you just delete a 0 and it will, at worst, add that hour to the next hour. With this current approach I end up just entering an 0 and the energy actually gained goes untracked forever

Maybe the solution with the approach we have is to somehow set it to half the next value and halve that next value as well

[simple olive]

Best solution would to be of course to have a not flaky sensor 😄

What is the sensor / integration? Maybe you can make a request to fix it there.

[thin tusk]

I am having the same struggles and am looking into this sollution now: https://www.libe.net/en/home-assistant-statistic

Essentially rebuilding the sum from the energy dashboard sensors through the sqlite browser addon