Mystery drop in SOC overnight....

Post Reply
calum
Posts: 398
Joined: Fri Feb 24, 2023 11:00 am
Location: Stockport

Hi everyone

Scratching my head about this.

Our system is as follows: H1 3.7, x6 HV2600 batteries, 14x400W panels, all installed last November.
Firmware versions as reported by the FoxESS app: Master 1.48, Slave 1.02, Manager 1.50
Inverter is connected via LAN cable to my home network and I use the Modbus variant of the StealthChesnut integration for HomeAssistant to monitor system performance.
We are on the Octopus Go tariff.

Since install I have had the system set to force charge between 00.30 and 04.30 since the solar gain was usually only enough to slow down the rate of discharge. Now we're getting a bit more sun I've been experimenting with charging for only part of that 4 hour window, to give room to capture solar gain the following day (and ultimately, reduce my electricity bill!)

However when I do this I'm seeing significant drops in the SoC when the system is meant to be neither charging or discharging. For instance last night I set the charging up as follows in the app:
Period 1: 00.30-03.00
Enable Force Charge = Yes
Charge from Grid = No
Period 2: 03.00-04.30
Enable Force Charge = Yes
Charge from Grid = Yes

In the screenshot below you can see the battery charge/discharge power (green line) stops following the house load (red line) at 00.30 as expected, although it maintains a flat discharge level of around 120W during this time (I assume this is a monitoring current from the BMS?)
At 03.00 the battery starts charging from the grid, again as expected. However what I don't understand is why the SoC (purple line, which is on the right axis, ignore that it says kW!) drops so steeply from just after 02.00 until the grid charging starts at 03.00.

Image

I can't get the "feed in power" trace to plot properly on the same graph for some reason, but I've looked at it and it's also zero at this time (meaning the battery is not discharging to the grid). I have also checked it for yesterday afternoon when the battery was full and we were returning power to the grid, and it looks as you'd expect. So as far as I can tell, the inverter is reading the current flows correctly.

The battery voltage trace for the same period is unremarkable, in that it's stable at ~316-317V from 00.30-03.00, when it increases in line with the charging curve, before dropping back down to about 321V when charging finishes, and slowly dropping thereafter - ie the voltage behaviour looks normal as well.
I've had a few other weird incidents where the BMS appears to have got 'stuck' somehow, so I'm wondering if this is "just" a BMS firmware bug - although it would be a worrying one if it can lose that amount of indicated SoC with minimal battery discharge indicated.

Any insights welcomed!
Attachments
Screenshot 2023-02-24 111854.png
H1-3.7 / 6xHV2600 / 14x400W / RS485 Modbus->HA
FoxESS Modbus HA Integration
Contact Fox here
xBoris
Posts: 5
Joined: Fri Oct 14, 2022 12:51 pm

I saw this sort of behaviour last summer. Consistently, 90mins into a "force charge" window, I would get about 90mins of rapid SoC loss without any significant power draw.

I escalated the issue to Fox in China over a few months and ended up on a series of unreleased/beta firmware versions for the BMS and batteries. Fox claimed that the issue was known and would be fixed via firmware.

Unfortunately, by the time a fixed firmware was available, we were deep into winter any I couldn't naturally validate the fix. I tried to reproduce the problem a couple of times and I didn't see any SoC drops. I also noticed that they seemed to have far less vampire draw during "force charge."

I'm still waiting for a better weather before further testing.

I believe that the new BMS/battery versions released a couple of weeks back are the final versions of what I have been using.

If yours are HV2600 V1's like mine, it might be worth a shot (HV2600_master_V1.015 and HV2600_slave_V1.11).
calum
Posts: 398
Joined: Fri Feb 24, 2023 11:00 am
Location: Stockport

Thanks. My HV2600s have "Ver. D" on a label on the front of them, not sure what version that is. I'm currently using the ethernet port for the TCP Modbus Home Assistant integration and I undertand the new firmware breaks that, so I need to look into a different way to achieve the same thing before I go asking to be upgraded. It does seem more like a firmware bug than an actual physical problem though.
H1-3.7 / 6xHV2600 / 14x400W / RS485 Modbus->HA
FoxESS Modbus HA Integration
Contact Fox here
xBoris
Posts: 5
Joined: Fri Oct 14, 2022 12:51 pm

Mine are version D purchased before V2 existed and therefore V1's. I believe the V2s are clearly labelled as such.

Its specifically the BMS and Battery firmware that helped me. I'm still running the old pre 1.50 inverter firmware myself (Master 1.48 / Slave 1.02 / Manager 1.47).

With that being said, if you ask Fox or your installer to update the battery, its entirely possible they will just update everything.
calum
Posts: 398
Joined: Fri Feb 24, 2023 11:00 am
Location: Stockport

My installers are pretty switched on about this sort of thing, I think asking them to only update the battery firmware would be the way to go. Thanks for the steer!
H1-3.7 / 6xHV2600 / 14x400W / RS485 Modbus->HA
FoxESS Modbus HA Integration
Contact Fox here
richt
Posts: 7
Joined: Mon Apr 10, 2023 9:39 am

Did you have any luck getting the SoC drop sorted after an update?
I have mine setup to dynamically top up between 1230 and 0430, and during this time set the 'house load' to 0 if no top up is needed.
I can repeatably watch the SoC drop by 20 - 25% each night during this idle period.
It's a similar story when charging - it will jump from 80% to 99% in the space of a minute.
Also using the LAN ModBus connection, so haven't plugged in the WiFi stick for months for fear of an update breaking things!

Some additional info:
A 99% to 20% discharge test on my 10.4KWh pack only yields 6.4KWh, so the SoC drop when idle actually supports the theory that the packs only have a 'real' capacity of just over 8KWh rather than 10.4KWh.
Either that, or they are way out of balance (something only the installer account can show IIUC).
SoC_Drop.png
H1-3.7 / 4xHV2600 / 10x390W / LAN Modbus->HA
G4RNI
Posts: 18
Joined: Wed Jan 25, 2023 12:56 pm

Last winter (2021 to 2022) when my system was new and I was as green as grass trying to make friends with it, I'd see my battery reserve fall by around 20 to 30% overnight. The system would also regularly top up from grid, drawing around 2kWh overnight. I then had 8 x HV2600 (20.48kWh they say) which I've now reduced to 7 on the advice of Fox techs.
I checked my house quiescent draw and actually consumed about 30W. The inverter wanted about 30W too "heartbeat". The fridge freezer would kick in for 20 minutes or so every couple of hours and it draws around 90W. So my average draw was a maximum of, let's up the figures and say 100W. In other words, 1kWh should last about 10 hours overnight but I was finding nowhere near that. Just to confirm, I was "losing" around 3 to 4 kWh into the mystic ether somewhere then topping up from the grid.

This is what I did: I went to the BMS at bedtime and switched it off. My consumption immediately dropped and the missing kWh were missing no more. I drew well under 1kWh overnight, from the grid and my batteries now lasted far longer.

My batteries weren't in the most agreeable place, in the loft on a hugely strengthened platform where once upon a time the water tanks lived and so the overnight temperature would drop to around freezing point. Perhaps this was the cause of the loss? Were the batteries trying like fury to constantly rebalance? You can tell when they're doing that by feeling their front panels, just around the painted spider webs. That is where the balancing load resistors live and when they're balancing, the panel gets warm to the touch. I admit, I didn't go up to check but perhaps they also have some inbuilt low temperature, self heating via the above dummy loads protection? If that's the case (and it may not be) then that's lost juice. I already know from my days as a comms & instrumentation tech just how inefficient batteries can become when cold. I've seen emergency backup banks lucky to provide half of their claimed capacity in cold situations requiring massive over engineering of banks.

I now have my second stack of batteries and new V2 BMS in the hall cupboard, along with the switching I've detailed elsewhere and my losses have virtually gone. Perhaps because the batteries are now at overnight room temperature, even in winter, around say 14°C???

I'll know better about the original batteries once the cold weather returns. Perhaps they're just grossly inefficient, being cold? Perhaps reducing the stack to 7 from 8 bats has improved efficiency enormously? Perhaps the V2 BMS is better at its job?
calum
Posts: 398
Joined: Fri Feb 24, 2023 11:00 am
Location: Stockport

I'm still trying to get the BMS and battery firmware updated, since I'm on 1.07/1.05, vs the versions reported higher up in this thread. I suspect at least part of this is a BMS firmware issue but until I get the latest I can't say for definite. I am seeing, best I can work it out, around 9-10kWh capacity, when I ought to have closer to 13 (6xHV2600 and a 15% Min SoC).

Regarding balancing, FWIW the BMS reports Cell mV High and Cell mV Low, as you'd expect these are both around 3.3V and almost always within 10mV of each other. Hopefully this indicates a fairly well balanced bank.
H1-3.7 / 6xHV2600 / 14x400W / RS485 Modbus->HA
FoxESS Modbus HA Integration
Contact Fox here
richt
Posts: 7
Joined: Mon Apr 10, 2023 9:39 am

I've figured out my issue...
I changed the Min SoC to 10% today and let everything discharge to cutoff.
I've then checked the voltage on each HV2600 and can see that 2 are slightly higher than the others.
Reconfiguring to 2xHV2600 rather than 4 and reconnecting to the inverter shows 30% SoC (previously 10%).
So it seems that the packs can only balance within one HV2600 and not across multiple packs.

I'm currently discharging those 2 to 10% and will see if things get a bit better balanced after tonight's off-peak charge!
H1-3.7 / 4xHV2600 / 10x390W / LAN Modbus->HA
richt
Posts: 7
Joined: Mon Apr 10, 2023 9:39 am

Looks like I was wrong.
Despite top balancing the packs, I still get a 20-25% drop when idle, and can only extract 7.1KWh from 99% to 10% cutoff (on a 10.4KWh pack).
Any suggestions for how to get a warranty claim started, or previous experiences?!
H1-3.7 / 4xHV2600 / 10x390W / LAN Modbus->HA
calum
Posts: 398
Joined: Fri Feb 24, 2023 11:00 am
Location: Stockport

What firmware are your batteries and BMS on? If not the latest then that would be the next troubleshooting step.

(I'm also seeing drops and sudden increases in SoC, and I'm working on getting the BMS and battery firmware updated, which is proving a lot trickier than I'd expected).
H1-3.7 / 6xHV2600 / 14x400W / RS485 Modbus->HA
FoxESS Modbus HA Integration
Contact Fox here
richt
Posts: 7
Joined: Mon Apr 10, 2023 9:39 am

Master : 1.30, Slave : 1.02, Manager : 1.50

If you find a way to get updated, please let me know :)
H1-3.7 / 4xHV2600 / 10x390W / LAN Modbus->HA
calum
Posts: 398
Joined: Fri Feb 24, 2023 11:00 am
Location: Stockport

Under discussion over here...
H1-3.7 / 6xHV2600 / 14x400W / RS485 Modbus->HA
FoxESS Modbus HA Integration
Contact Fox here
richt
Posts: 7
Joined: Mon Apr 10, 2023 9:39 am

Thanks for the link. The WiFi stick route sounds like a good fall back plan.
I got a response from my installer (who received a response from Fox), and they have upgraded the battery FW:
I have checked the battery version is 1.0, which has the problem of SOC imbalance. I have upgraded the battery, please check it again.
So far, the SoC jump / drop hasn't happened yet but only got down to 40% last night.
H1-3.7 / 4xHV2600 / 10x390W / LAN Modbus->HA
calum
Posts: 398
Joined: Fri Feb 24, 2023 11:00 am
Location: Stockport

I'm going to try and do the f/w upgrade via the spare WiFi stick route this evening, I'll report back on how we do with weird SoC changes after that /crosses fingers
H1-3.7 / 6xHV2600 / 14x400W / RS485 Modbus->HA
FoxESS Modbus HA Integration
Contact Fox here
richt
Posts: 7
Joined: Mon Apr 10, 2023 9:39 am

FWIW, my experience after the upgrade:
The SoC is more stable during use; however when the power draw from the battery sits at zero for > 1.5 hours, the SoC will adjust (dropped by ~10%).
I also have the same behaviour as before when getting to full charge: SoC rises normally, then jumps by ~20% to full.

SoC went from 33% to 23% between 0030 and 0430 (Grid only mode for Octopus Go), then during the day went from 23% to 83% before jumping to 99% with just 6.17KWh added.
The rise from 23% to 83% is correct in that it matches the 60% charge added; however the jumps at high / low SoC mean I can only get about 60-70% of the advertised capacity from the pack.
SoC_Drop2.png
H1-3.7 / 4xHV2600 / 10x390W / LAN Modbus->HA
calum
Posts: 398
Joined: Fri Feb 24, 2023 11:00 am
Location: Stockport

You may want to do some full -> empty -> full cycles to help the BMS "learn" about your new battery with the "better understanding" it has from the new firmware. If that sounds hand wavy, it's because it is, a bit, but it has paid dividends in my case. To be fair, I have yet to leave the battery without any load on it overnight, but I do at least appear to be getting the right amount of charge when going from Min SoC to Max SoC.
H1-3.7 / 6xHV2600 / 14x400W / RS485 Modbus->HA
FoxESS Modbus HA Integration
Contact Fox here
richt
Posts: 7
Joined: Mon Apr 10, 2023 9:39 am

Thanks - the hand-wavy theory makes sense, as it needs to know the full / empty capacity to calibrate the SoC.
I've done about 6 full cycles: Charge to 99%, watch battery input drop to 0KW, then discharge to empty when the 0.8KW emergency charging kicks in to top it back up.
It still plays silly beggars, jumping by 20% at the top / bottom (or idle).

Fox have replied again saying they will consult their team and arrange a replacement if it continues.
H1-3.7 / 4xHV2600 / 10x390W / LAN Modbus->HA
calum
Posts: 398
Joined: Fri Feb 24, 2023 11:00 am
Location: Stockport

We went out today leaving the battery at 12% SoC, and came back to find it at 100%. Last 10% or so was a near vertical line on the trace, but as I'm having what look like some balancing issues with my batteries, I'm going to try to rebalance them which will hopefully help with both.
H1-3.7 / 6xHV2600 / 14x400W / RS485 Modbus->HA
FoxESS Modbus HA Integration
Contact Fox here
Fursty Ferret
Posts: 3
Joined: Tue May 16, 2023 11:14 am

@richt - did you ever get to the bottom of this? I've seen some wild swings in my battery capacity over the last few weeks with it tumbling from 90% to 20% in the overnight force-charge interval, and then only recharging to 85% despite 3kW of solar available.

Have tried the techniques suggested elsewhere (4A trickle charge, full discharge and refill etc) without much success. Inverter is on latest firmware though don't know if there's a separate firmware for the BMS and batteries as I don't have agent access. I've checked the individual battery voltages and they're exactly the same (53.5V on each of the three batteries).
calum
Posts: 398
Joined: Fri Feb 24, 2023 11:00 am
Location: Stockport

don't know if there's a separate firmware for the BMS and batteries
There is and you should definitely get it updated to the latest - I saw big improvements when I did this. You can open a ticket with Fox UK if you need to.

The battery voltages can mask cell level voltage differences which the BMS does track, so it comes back to getting the FW for that updated.
H1-3.7 / 6xHV2600 / 14x400W / RS485 Modbus->HA
FoxESS Modbus HA Integration
Contact Fox here
Post Reply