Rapid Change in Cell Voltage

[Nightguest]

I'm a little stumped here, so I'm wondering if anyone has any advice for me.
I've got 2 batteries, an SOK 48/100 and a DIY 280Ah one, and until now they've been totally fine.

However, recently I've been getting a "high cell voltage" alarm in my CerboGX, and I've noticed that in both my SOK battery and my DIY battery, 1 cell seems to go "all over the place.

What I noticed today:

• I've got about 1400W going into the battery.
• According to both the CerboGX and the BMS screen, cell 4 jumps between 3.24V, 3.30V. 3.42V, 3.65V, 3.9V (slightly alarmingly), then drop back down to 3.245V, then back up.
• However, when I discharge the battery pack, that cell seems to behave exactly like all the others. For example, I've got my regular loads, then turn the washing machine on, comes up to about 2400W being pulled, and all the cells behave the same, with no cell going out of balance or
• The BMS does "try" to manage this, I notice that every time the voltage spikes, the entire module/battery pack is taken offline (again, shows in CerboGX, 1 module online, 1 offline), and then when the voltage is below 3.6V (about 1s later) the module turns back online. (I can hear the BMS clicking on/off/on/off when this state triggers).

In my SOK battery, it seems that cell 16 does something similar, but to a lesser extent than in my DIY battery (it will go from 3.25 to 3.6V, then down again).

So, I'm wondering if this is a case of a defective cell that needs replacing, no big deal if yes, just slightly annoying. But, more importantly, if yes, is this potentially dangerous?

Or, is this something else that I'm missing entirely?

 

[Steve_S]

There are a few possibilities at play here:
1) Runner cells which can "Race UP or Drop Fast" very fast once outside of the working voltage (above 3.450 or below 2.900). These are usually apparent right from the start.
2) a Loose connection can seriously skew readings. Check all busbars & Sense/balance leads to make sure everything is clean & tight.
3) a Charge profile that pushes "just that bit too much" can trigger runners (going high) and the only fix is to make a slightly more moderate charge profile. Most often, people try to push the envelope of what should be used to "squeeze" more out. Kinda like when the gas pump kicks off when filling your car and you push in another Litre or so to "top it off". That does NOT work with batteries... causes more problems.
4) In some more rare instances, a "borderline" cell may have been used and over time (and too much push) which has weakened it a bit.

Keep in mind two FACTS.
LFP like all other battery chemistries have 2 Voltage Ranges. The "Allowable" range which does not harm or damage LFP is from 2.500 to 3.650. This is the safe zone for the cells to be "within". The "Working" range is what the LFP cells are designed to operate within on a daily & normal usage. This Range is from 3.000 to 3.400 volts per cell. A-Typically this is safe to charge up to 3.450-3.475 because LFP always settles close to 3.400 after full charge, and can be safely discharged without issue down to 2.800. Below 2.800 runner cells will drop fast to LVD state.
NOTE that Factory Capacity Tests on LFP are within the Woking Range which is actually where the rated AH comes from, hence why Nominal Voltage for LFP is 3.200 Volts (aka 50% SOC).

A good Moderate Charge Profile that works with Varied Capacity Packs (including 100/175 & 280 AH packs) within a Bank.
This is using Midnite Classic SCC but setting options are very similar for other SCC's. There is also BMS Tweaks for Standard JKBMS with Active Balancing.

Steve_S-Tech: My Final Config for Midnite Classic SCC & JKBMS' w/ Large Bank (works amazingly well) Includes Charge Profile & BMS Config

INTRO: I have debated posting this for a while until I got everything tweaked in and working as best as possible. The End Result is Gobsmackingly Good ! I cannot speak for other SCC and such as each is different & unique in their own ways BUT this may likely be a Good Reference. Civility &...

diysolarforum.com

Hope it Helps, Good Luck.

This may help as a quick reference...

 

[400bird]

How quickly does it bounce around?

Sounds like a bad connection in the wiring from that cell to the BMS.

 

[DIYrich]

Voltage can bounce around as the bms disconnects and reconnects the battery.

Try the slow way of top balancing. Start at 53.0v, and let it settle. Drop the amps to .05c, and increase in 0.2v increments. Let settle, and repeat.

I had to do this with a 280ah battery. Voltage differential was 0.25v between high and low cell. When the BMS disconnected the battery, the low voltage cells dropped faster between cycles than they charged. I needed to keep the charge voltage going while it balanced (don't let the bms disconnect the battery).

I charged to the voltage where the high cell was 3.640v. I let it balance to where the high cell was 3.630v, and increased charge voltage by 0.10v. It took a day. Slow at first. Once it got to 57.0v, it went faster. Eventually got to 57.6v (3.60v per cell), and I let it finish balancing to 0.005v difference.

 

[Nightguest]

Thanks for the quick help here, and I've managed to solve the problem.

It turns out that one of the busbars on cell 4 was really loose (all of the others were perfectly tight, just that 1 had become loose). I have no idea how, but, tightening it up solved everything.
*Honestly, I totally forgot that this could have been the case!!

So, thanks for the really quick advice and the info shared here, you lot rock!

*I am curious, do you lot do checks on how tight the nuts are screwed in every few months, or is that not something that normally needs to be done?

 

[Steve_S]

You can't really use anything like LocTite because it's non-conductive. Cell terminals that require "Screws" typically have Lock Washers (split or Serrated), if welded posts Serrated Nuts or nuts with lock washers at least. If properly torqued they should not loosen. Unfortunately, the torque settings are tricky depending on the type of terminal, welded posts versus tapped holes or terminal block.