Describe the bug
I've had many instances recently where the plan isn't economic, it's going to cost more than doing nothing and there are clearly better ways to schedule charging.
I've looked at the config, tried tweaking a few things, not made any real difference.
Example today, files attached: I had predbat in monitor mode (because of the problem) and it was forecasting the rest of the day to cost -£0.88, which is reasonable; it's 3:30pm, we've been on PV since about 7am and the battery is full. I switched to 'control charge and discharge' mode and checked the plan. Charging slots all over the place, some clearly in places where it could have waited an hour for cheaper rates, and an expected cost for today of +£2.00. Predbat's own modelling is telling me that its plan is going to cost me way more than doing nothing!
The only thing that has changed, is that I added more battery capacity on Tuesday. I set all the paramters correctly for the additional capacity - battery rate was 3000, now 5500, battery max soc was 10.2, now 20.4 and everything appears to be operating with the new capacity. It's now Saturday. I've tried setting days_previous to 1 so that it's not looking at any history with the smaller battery, just in case there's something that doesn't cope with the difference in the history, but even with it looking back only to yesterday, the plans are still really strange. Load and PV figures on the plan all look reasonable, it's just the decisions it's making that don't make sense in relation to the data.
Predbat was fine previously for a year plus. I just can't seem to track down what's causing this behaviour. I'd really appreciate some expert help tracking this one down! Hopefully a more knowedgeable eye will spot something I didn't.
Monitor mode forecast -£0.88 today (£2.12 tomorrow):
Control C&D mode plan (days_previous = 1) +£2.00 (£4.21 tomorrow)
When I look at this plan, it just makes no sense to me - for example, it's charging at 20:30 and again at 21:30, yet we know that there's plenty enough in the battery to wait. It clearly could wait until 22:30 and do it for less. That's if it were appropriate to charge at all, which the monitor mode projection of 'let it be' suggests it isn't.
Expected behaviour
The predbat plan should be more economic that simply leaving the system alone, or its plan should be to just let the system run. It was that way until last week... what's broken in my system?
Predbat version
8.44.11 (problem the same back to at least 8.44.5)
Environment details
nipar44 docker image / HA container, HA 2026.7.2 (same problem with 2026.6)
Solax inverter, modbus mode 8 control
I removed the entire predbat folder structure and pulled the image again into an empty folder today in case something was corrupted. No change.
Predbat files
predbat_debug.yaml.txt
predbat_plan.html
predbat.log
apps.yaml.txt
Describe the bug
I've had many instances recently where the plan isn't economic, it's going to cost more than doing nothing and there are clearly better ways to schedule charging.
I've looked at the config, tried tweaking a few things, not made any real difference.
Example today, files attached: I had predbat in monitor mode (because of the problem) and it was forecasting the rest of the day to cost -£0.88, which is reasonable; it's 3:30pm, we've been on PV since about 7am and the battery is full. I switched to 'control charge and discharge' mode and checked the plan. Charging slots all over the place, some clearly in places where it could have waited an hour for cheaper rates, and an expected cost for today of +£2.00. Predbat's own modelling is telling me that its plan is going to cost me way more than doing nothing!
The only thing that has changed, is that I added more battery capacity on Tuesday. I set all the paramters correctly for the additional capacity - battery rate was 3000, now 5500, battery max soc was 10.2, now 20.4 and everything appears to be operating with the new capacity. It's now Saturday. I've tried setting days_previous to 1 so that it's not looking at any history with the smaller battery, just in case there's something that doesn't cope with the difference in the history, but even with it looking back only to yesterday, the plans are still really strange. Load and PV figures on the plan all look reasonable, it's just the decisions it's making that don't make sense in relation to the data.
Predbat was fine previously for a year plus. I just can't seem to track down what's causing this behaviour. I'd really appreciate some expert help tracking this one down! Hopefully a more knowedgeable eye will spot something I didn't.
Monitor mode forecast -£0.88 today (£2.12 tomorrow):
Control C&D mode plan (days_previous = 1) +£2.00 (£4.21 tomorrow)
When I look at this plan, it just makes no sense to me - for example, it's charging at 20:30 and again at 21:30, yet we know that there's plenty enough in the battery to wait. It clearly could wait until 22:30 and do it for less. That's if it were appropriate to charge at all, which the monitor mode projection of 'let it be' suggests it isn't.
Expected behaviour
The predbat plan should be more economic that simply leaving the system alone, or its plan should be to just let the system run. It was that way until last week... what's broken in my system?
Predbat version
8.44.11 (problem the same back to at least 8.44.5)
Environment details
nipar44 docker image / HA container, HA 2026.7.2 (same problem with 2026.6)
Solax inverter, modbus mode 8 control
I removed the entire predbat folder structure and pulled the image again into an empty folder today in case something was corrupted. No change.
Predbat files
predbat_debug.yaml.txt
predbat_plan.html
predbat.log
apps.yaml.txt