How to identify that predicted high load in the plan and chart session?

[ppmt]

Hello,
For the last few days I noticed that the load is always high nearly for the full day but then as the time goes by it readjust itself.

So I tried to identify what load could cause that. On Thursdays and Fridays I have 2 big oven that are on in the evening but at most they consumed about 12kw through each night. So that doesn't explain the peak consumption reported in the LoadML charts.

I may have charged my cars on one of these days but the sensor for them are listed in the car_charging_energy.
An in any case I have never had my cars plugged in for such long period.

So what files could me help identify why there is such a high load? I looked through the debug files but I can't really make sense of it.
I want to learn how to debug it so it is more guidance that I am after rather than someone doing it for me :)

[johnwb87]

We are all learning about LoadML. It appears to be very sensitive to the sensors that it relies on. Those being load and pv. Check those first for LoadML issues.

[gcoan]

switch.predbat_calculate_inday_adjustment is set to true. Is this what you are referring to?

that is probably your issue https://springfall2008.github.io/batpred/load-ml/#history-accumulation-and-the-database

[ppmt]

Two things to look at to see if the car energy is being filtered.

One, chart the LoadML actual with your main house load sensor and see how they track each other.

Second go to the Predbat WebUI, go to log, click on all and the out ML in the search. There should be a 2 log messages each 30 minute update run that reference car charging subtraction.

yes I can see these lines. What I don't know is what is substracted. I defined 2 sensors, one for my main charger and one for the granny charger .

53559	2026-03-27 11:46:26.917953: ML Component: Car charging subtraction applied, 4033 data points after subtraction
51740	2026-03-27 11:16:22.268168: ML Component: Car charging subtraction applied, 4033 data points after subtraction
49745	2026-03-27 10:45:16.105782: ML Component: Car charging subtraction applied, 4033 data points after subtraction
49602	2026-03-27 10:44:13.387590: ML Component: Car charging subtraction applied, 4033 data points after subtraction
47804	2026-03-27 10:14:08.810439: ML Component: Car charging subtraction applied, 4033 data points after subtraction

The PV energy goes into the model and is used for training to see if any patterns in energy use are correlated with it. I think Trefor’s idea was that certain solar owners have become used to using more electricity on sunny days so letting the ML model look for that pattern is a good thing.

[ppmt]

switch.predbat_calculate_inday_adjustment is set to true. Is this what you are referring to?

that is probably your issue https://springfall2008.github.io/batpred/load-ml/#history-accumulation-and-the-database

Ah!!! I will disable it.

The documentation is great and very detailed...may be sometime too detailed and you miss that little line that can change everything.

There is so much that Predbat can do that it gives me a lot of respect for people like you that are able to keep on top of it.

[gcoan]

The documentation is great and very detailed...may be sometime too detailed and you miss that little line that can change everything.

There is so much that Predbat can do that it gives me a lot of respect for people like you that are able to keep on top of it.

yes I'm just updating the docs now to make this even more prominent. Previously it was even less clear, it just said 'for best results turn off inday adjustment' but for lots of people they saw their charts escalate too rapidly

There is so much that Predbat can do that it gives me a lot of respect for people like you that are able to keep on top of it.

some of the time I end up reading the code to work out what it does and then write the documentation ...

[ppmt]

some of the time I end up reading the code to work out what it does and then write the documentation ...

I mean....I rest my case 🙇‍♂️