Li Ion batteries - if Predbat plans 100% charge overnight does this mean battery full?

[Javert1969]

Testing predbat over multiple nights I noted that because I am on the IOG tariff with a long overnight charging window, Predbat usually decides to charge the battery to 100% overnight, but does 100% mean "full"?

I noticed that Predbat always stops the charge proactively as soon as soc 100% is reported by the battery and then floats the battery until the next mode change. Predbat stopped the "charge_start" and changed to "charge freeze" as soon as 100% was reported but the battery was not completely full to the brim.

With Li Ion batteries, this does not mean the battery is full - at least this is the case with Alpha ESS inverters for example. The battery will continue to charge for some more time until it actually refuses to accept any charge. In fact, Alpha ESS has a Modbus available flag "battery full" which is exposed by the integration and which tells you when the battery is actually full. In my testing, "battery full" was never reported overnight - it was only reported during the day when in demand mode and there is plenty of solar. On the other hand if I charge manually overnight, "Battery Full" is reported and the battery floats automatically at battery full until the end of the programmed charge - no intervention necessary this happens as long as the timer is running.

So my question - is it "by design" that Predbat behaves in this way or do I have an issue with my config? Probably you already know all this and this is done on purpose this way for some reason?

I tested this with and without low power settings. I have set up a charge curve to trick the system into thinking that charging is slower at higher levels (even though actually my battery will accept almost full charge until after it reports 100%). I also tried messing with the "low_power_mode_margin" but this didn't help either.

Maybe there is another setting I have not found yet that will help with this? When charging the battery to 100% overnight, my working assumption would be that I want to charge it to completely full until it actually will not take any more charge.

Obviously I can work around this outside of Predbat, to force a top up charge near the end of the cheap period but I'm wondering if there is a reason it's been designed this way?

This is not an issue most of the time but for example on the darkest mid winter day, this could make the difference between running out of battery in the end of the evening or not?

[gcoan]

Predbat will as you have identified stop force charging the battery when the SoC is reported as 100%. There are some reasons for this that I can't remember off hand, maybe something to do with not overcharging, IDK.

So if your inverter isn't really full when it reports 100% SoC then that could be a problem. It's probably not going to be much different but I'd suggest you'd probably need to create a template sensor that is driven from the battery SoC level and the separate 'full' indicator - e.g. reports 99% full if SoC is 100% and battery isn't reporting full.
Something like that would work.

[WyndStryke]

I'd suggest you'd probably need to create a template sensor that is driven from the battery SoC level and the separate 'full' indicator - e.g. reports 99% full if SoC is 100% and battery isn't reporting full.

That's what I was about to suggest.

I don't know anything about AlphaESS batteries in particular, but what I would want to know is if the 'early 100%' is enough to do battery pack balancing and battery voltage calibration, or if it needs to go to 'true 100%'.

If 'early 100%' is enough then I would suggest sticking with the current setup, and don't try to reach true 100%. If however, you need to hit 'true 100%' periodically to retain voltage calibration then the modified sensor would work (but typically you only need to do voltage calibration once every week or so).

My Fox system, and I suspect most others, treats 100% as a hard-stop. I suspect the reason that the AlphaESS has this 'early 100%' mechanism is to avoid the extra battery degradation you get at the very top of the range.

[Javert1969]

Not sure about this because if you charge the battery using the Alpha ESS app with 100% as the target, it carries on charging until full.

Also - from what I can see, there is very little charge curve at all until the SOC reports 100% - it's only at reported 100% SOC that the charge curve starts to kick in.

In any case, in the meantime I have already implemented what you suggest. What I am finding is that in low power mode, this sometimes results in the battery getting completely full when Predbat charges to 100%, but sometimes it just misses it - carries on charging for some time after 100% but then reaches the time limit.

[WyndStryke]

Also - from what I can see, there is very little charge curve at all until the SOC reports 100% - it's only at reported 100% SOC that the charge curve starts to kick in.

In any case, in the meantime I have already implemented what you suggest. What I am finding is that in low power mode, this sometimes results in the battery getting completely full when Predbat charges to 100%, but sometimes it just misses it - carries on charging for some time after 100% but then reaches the time limit.

What I would suggest is to manually enter a charge curve in predbat which has a significant slowdown for 98, 99 %, so that it is forced to reserve extra time. Often this extra time would be unnecessary, but it would mean that the occasions when it charges slowly are covered.

What I personally find is that from 90% upwards, charging times on my system can be quite unpredictable. Sometimes it will take extra time (presumably to do balancing & voltage calibration), and other times it will just skip several % at a time (presumably because the BMS decides it is unnecessary). Today was the latter, it pretty much just skipped over the higher SoC range.

[Javert1969]

I have added a charge, curve, and what I have also done which I will test tonight is to further adapt my SOC indicator. Based on my observations of the amount of charge that the battery takes after the SOC reads 100%, it often takes another 0.5 to 0.75KWH before it reads full, which represents more than 3% of the battery. As such, working assumption is that 100% actually means 97-98%.

Therefore I have adapted the Predbat SOC indicated, withhelp from Claude.ai, to do a scaled approach - the SOC reported to Predbat gradually scales back after 50% to a lower value, reaching 98% when the Alpha SOC is 100%. I can tweak the parameters and scaling factors with experiments.

If this doesn't work, I will intervene in the power to notch up the charge power slightly from what is ordered by Predbat with a cap at the inverter max.

Draft code is
template:

• sensor:
  • name: "Battery SoC Adjusted"
    unique_id: battery_soc_adjusted
    unit_of_measurement: "%"
    device_class: battery
    state_class: measurement
    state: >
    {% set soc = states('sensor.alphaess_soc_battery') | float(0) %}
    {% if soc >= 100 and is_state('sensor.alphaess_battery_full', 'True') %}
    {{ 100 }}
    {% elif soc > 50 %}
    {{ (50 + (soc - 50) * 0.96) | round(0,'floor') }}
    {% else %}
    {{ soc | round(0,'floor') }}
    {% endif %}
    availability: >
    {{ states('sensor.alphaess_soc_battery') not in ['unknown', 'unavailable'] }}

[gcoan]

Probably also worth saying that for all inverter batteries, 100% is never totally full anyway, and neither is empty totally empty. The inverter should operate the battery within an 'almost empty' to 'almost full' range as the cells don't like being totally exhausted or filled