Predicted load today seems to have changed to 0 for the rest of the day.

[jambees]

Describe the bug
From around 12pm today the load kWh shown on the plan is 0.0. I've done some checking in the config file and the load_today config entry is set to sensor.load_energy_today which is my fox ess modbus sensor. That appears to still have data in it that is increasing for the day. Also strangely, if I go to the apps tab in predbat addon and click the load today sensor I get a sawtooth graph showing load_today value increasing each day for the last 7 days and resetting at midnight. So I can't see why my predicted load has now gone to 0.

Expected behaviour
Predicted load showing non zero values that reflect a reasonable guess of the load based on historical values

Predbat version
v8.32.14

Environment details

• Fox ESS inverter and Fox ESS energy cube battery
• Standard HAOS

Screenshots

Predbat plan:

predbat_plan.html

Load energy today graph:

Log file
Log file attached. I saw some errors about watchlist that were possibly due to a bad configuration - I've commented out the two watchlist items from the config file for octopus saving sessions and octopus car charging now.

predbat (3).log

Predbat debug yaml file

Debug yaml file attached

predbat_debug (1).txt

[gcoan]

I get a sawtooth graph showing load_today value increasing each day for the last 7 days and resetting at midnight. So I can't see why my predicted load has now gone to 0.

The sawtooth is largely irrelevant as you have days_previous set to 7 so predbat is only taking your load from 7 days ago rather than averaging multiple days

But your load sensor from 7 days ago has got house load in it, so that's not the problem

What is the problem is shown in the logfile:

2026-02-02 12:50:02.590509: Historical load totals for days [7] are [6.24]kWh, minimum value 6.24kWh
2026-02-02 12:50:02.591325: Historical days now [7] weight [1]
2026-02-02 12:50:02.610272: Today's load divergence 83.99%, in-day adjustment 91.87%, damping 0.95x, yesterday 100.0% today 84.79% blend 46.53%
2026-02-02 12:50:02.610340: Today's predicted so far 9.74kWh with 8.03kWh car/iBoost excluded, 5.28kWh import ignored, and 0kWh forecast extra.
2026-02-02 12:50:02.610367: Today's actual load so far 6.94kWh with 0.0kWh car/iBoost excluded, and 4.37kWh import ignored.

All of your house load from 7 days ago is being treated as car charging energy so is being ignored.

You need to look at the setup of your car charging load sensor, that is what is causing your house load to be ignored. Most likely it is not incrementing or resetting properly as it needs to for predbat https://springfall2008.github.io/batpred/apps-yaml/#car-charging-filtering

[jambees]

Hmm. So I had a look at the config and the car_charging_energy seems to point to a valid sensor which does increase and reset.

However this got me thinking - I was away visitng family last week which in itself I didn't think would impact the predicted load (might be a bit lower but I expected that). But the car charging while I was away is still being recorded in the sensor (which makes sense) but perhaps this is negatively impacting the load as it wasn't seen in the house load. I still can't get my head around that fully and and not quite sure how to fully understand that log excerpt you provided.

[jambees]

I had a read of holiday mode in the docs but I don't think that'll fix the issue. What might is to set days_previous to 1 and slowly increment over the next week back to 7 days if the away from home car charging is the problem with the predicted load.

[gcoan]

In general I recommend you set days_previous to multiple days, not just a single day, because then you get a better average load figure rather than just what the load was a week ago.

Can you share your house load sensor and car load sensor history from 6-8 days ago, that will probably reveal the problem. More recent days are not relevant.

[jambees]

House load used for predbat's load_today: config entry:

Car energy added sensor used for predbat's car_charging_energy config entry:

Both graphs show 24th Jan 06:00 to 29th Jan 09:00. I did a lot of driving on the 24th to the holiday destination with some charging along the way, arriving in the evening of 24th. Some slow 2kw granny charger charging on 26th and 27th.

[gcoan]

So there's your problem on the car charging energy, it is not behaving as Predbat requires it to.

On Sunday the sensor reduces in value and resets to zero several times through the day

On Monday the sensor doesn't reset to zero

Tuesday and Wednesday the sensor reduces in value

From the documentation https://springfall2008.github.io/batpred/apps-yaml/#car-charging-filtering

If you are getting erroneous house load predictions in your plan then check this setting and car_charging_energy are set correctly.

car_charging_energy - Set in apps.yaml to point to an entity which is the daily incrementing kWh data for the car charge

Try wrapping the car energy sensor in a daily resetting utility meter, that should stop the problem of your sensor reducing in value

[jambees]

Thank you for your analysis and advice.

I had already created a helper utility meter for lifetime energy added to use in the HA energy dashboard (as the one provided by the tesla integration was doing strange stuff with its resets in that dashboard) , so I have now created another utility meter using the total increasing helper sensor as it's source with a daily reset which I think will meet the Predbat sensor requirement.

I imagine that my predicted load has been slightly off for a long time then and should now be more accurate moving forward with this change.

[gcoan]

Thanks, I am pretty sure this will resolve it, but give it a couple of days to confirm.

You're not the first person who has been caught out by this, I have further extended the documentation to highlight the energy sensor requirements, and suggest creating a utility meter to provide a sensor as predbat requires