LFP memory effects thread

[Cpt Pat]

Here's a good paper describing the coulomb-counting method.

https://arxiv.org/ftp/arxiv/papers/1803/1803.10654.pdf

Referring to figure 2 therein, I can summarize a practical application of the method as:

At or near 25 degrees C (77 F):

1) To start, you need to determine where 100% SoC is. If you go over, you'll damage the batteries. I charge at 0.16C (I suggest anywhere between 0.10 and 0.20C) while monitoring the current for a taper to 0.05C. Monitor the per cell voltages during charging to assure they do not exceed 3.6 volts per cell each (manually balance the cell voltages with a shunt resistor where needed. I use a 2.7 ohm 10 watt resistor.).
2) Rest the battery in open-circuit condition for a few hours. Measure the open circuit voltage (OCV). In a 4S pack, the OCV should be above 13.37 volts with absolutely no load (besides your volt meter). Note: this voltage is "close enough." You're in the overcharge danger zone here where you're in the "charge knee" of the curve and any increase in voltage yields very little extra capacity. I disregard the few percent possible undercharge error that may exist with that voltage. Safety is more important than precise accuracy at this point.
3) If the resting OCV voltage is less that 13.37 volts, add a few more amp/hours of charge, then rest, measure, and repeat if needed.
4) Set your amp/hour counter/SoC meter (I use a Victron BVM-702) to "100%."

You now have your 100% set point. You can use your amp/hour counter to terminate charging by setting the threshold where you desire. I use 80%. Many of you may prefer 90%.

If you can't achieve an OCV resting voltage of 13.37 volts after charging to the current taper (0.05C) with the cell voltages at 3.6 volts during charging - you have weak or "memorized" cells. Discharge the batteries to 10% SoC and repeat the charging process. You should show an improvement in the resting OCV voltage with each repetition.

Verify the resting OCV voltage with the SoC display during resting periods with little or no load on the battery. You can find a good OCV versus SoC voltage graph within the page here: Marine lithium batteries in operation | Nordkyn Design. That page also has a good discussion about the memory effect. You should find that the displayed OCV very closely and repeatedly matches the displayed SoC, based on the graphs in the link above. If not, you may have to reduce the A/H capacity setting on the meter to reflect the reality of old/damaged/overrated cells.

Repeat this cycle to recalibrate around every 10 to 50 battery cycles. The recalibration process itself should erase the memory affects on moderately affected batteries.

The root of the memory effect evil is the use of voltage thresholds alone for determining charge termination during charging. That method causes progressive accumulation of errors. That's why I use coulomb-counting instead.

A voltage threshold can be safely used to determine where to terminate discharge. I use 12.7 volts, which equals about 15% to 20% SoC with my level of discharge currents (0.10C average). My batteries then rest to an OCV of 12.9 volts after the load is removed.

[OceanSeaSpray]

I am reposting a summary of what I had in the other thread for the benefit of switching to this dedicated - and hopefully less "noisy" - thread. Thanks to evm1024 for creating it.
This below was what started the memory effect discussion and, from now on, I will only discuss it here.

I have been running a very interesting fractional cycling experiment here for 4.5 years and now this bank stays in absorption for over an hour at the end and only reaches 70% SOC when I take it to the termination point of 14.0V and 0.03C.
Recharging the last 30% would take hours and hours at very low current. Basically, it shows very high internal resistance when the SOC approaches this point and it doesn't behave normally at all. Other then that, it performs fine. The internal resistance is generally a bit higher than when it was new, but this is normal with aging.

In order to assess capacity, I started recharging with the pack very close to the bottom, 2.95V/cell or so, and recharged continuously until the termination condition was hit while metering the current. I charged at about 0.7C initially and then the pack went into absorption for over an hour and the current stayed very low for most of that time. At the end, the apparent resistance of the pack was very high, about 125mOhms, but only for charging. Another curious thing I observed was that if I then suddenly discharged some capacity, I could put that straight back in at high current, but I would hit the wall again at the same point again after that.

After the charge, the cell voltage settled back down to a value abnormally low for fully recharged cells and in agreement with the charge supplied. This is a tell-tale of memory effects and what prompted me to formally measure the available capacity. When a 4S pack has been fully recharged, the voltage under light load should hold above 13.3V for about 0.2C. If this doesn't happen, there is an issue somewhere.

I hoped that the next full charge cycle after that would be easier and take me further, but it didn't.

The research indicates that memory effect is the result of failing to recharge completely and it becomes more and more noticeable with the number of aborted charge cycles, especially if they all stop in the same place. In my case, winter essentially prevents full recharging. I will have had well over a thousand of these incomplete charge cycles by now, stopping anywhere randomly. In summer, when the energy was available to fully charge, the charging regime used was obviously not enough to (fully?) overcome the cell memory in spite of the 3.5V/cell absorption voltage etc. As a result, it has been getting worse.

At this stage, this seems to suggest that charging should probably be adaptive and take into account the cycling history.

This is now the 5th winter for this bank and it is still operating fine because there is still more than enough capacity for routine daily cycling, but reserve capacity is down. If I hadn't taken the trouble of looking into it and measuring it, the situation wouldn't look much different than when the cells were new.

The objective now is erasing that memory and getting back to normal performance.

[OceanSeaSpray]

Originally Posted by Cpt Pat

Here's a good paper describing the coulomb-counting method.

https://arxiv.org/ftp/arxiv/papers/1803/1803.10654.pdf

Referring to figure 2 therein, I can summarize a practical application of the method as:

At or near 25 degrees C (77 F):

1) To start, you need to determine where 100% SoC is. If you go over, you'll damage the batteries. I charge at 0.16C (I suggest anywhere between 0.10 and 0.20C) while monitoring the current for a taper to 0.05C. Monitor the per cell voltages during charging to assure they do not exceed 3.6 volts per cell each (manually balance the cell voltages with a shunt resistor where needed. I use a 2.7 ohm 10 watt resistor.).
2) Rest the battery in open-circuit condition for a few hours. Measure the open circuit voltage (OCV). In a 4S pack, the OCV should be above 13.37 volts with absolutely no load (besides your volt meter). Note: this voltage is "close enough." You're in the overcharge danger zone here where you're in the "charge knee" of the curve and any increase in voltage yields very little extra capacity. I disregard the few percent possible undercharge error that may exist with that voltage. Safety is more important than precise accuracy at this point.
3) If the resting OCV voltage is less that 13.37 volts, add a few more amp/hours of charge, then rest, measure, and repeat if needed.

I think that the general idea is sound. If you charge LFP cells at any "charging" voltage (i.e. >=3.4V/cell) and keep going, the current eventually goes down to zero, which is undesirable because the cells have been overcharged by then. However, as long as the voltage is limited, there is no immediate "safety" issue with this.

Originally Posted by Cpt Pat

If you can't achieve an OCV resting voltage of 13.37 volts after charging to the current taper (0.05C) with the cell voltages at 3.6 volts during charging - you have weak or "memorized" cells. Discharge the batteries to 10% SoC and repeat the charging process. You should show an improvement in the resting OCV voltage with each repetition.

There may not/should not be any need to discharge. Just keep absorbing

Originally Posted by Cpt Pat

Verify the resting OCV voltage with the SoC display during resting periods with little or no load on the battery. You can find a good OCV versus SoC voltage graph within the page here: Marine lithium batteries in operation | Nordkyn Design. That page also has a good discussion about the memory effect. You should find that the displayed OCV very closely and repeatedly matches the displayed SoC, based on the graphs in the link above. If not, you may have to reduce the A/H capacity setting on the meter to reflect the reality of old/damaged/overrated cells.

Yes, the maximum "absorbable" capacity will eventually reflect the SOH of the cells. Note that you can get an early hint by comparing the difference left to reach nominal capacity and the SOC derived from the OCV. This also tells you how much more to absorb at CV, no need to go up a few Ah at a time.

Originally Posted by Cpt Pat

The root of the memory effect evil is the use of voltage thresholds alone for determining charge termination during charging. That method causes progressive accumulation of errors. That's why I use coulomb-counting instead.

That would be a circular argument, because standard CC relies on identifying termination based on voltage and residual current and this limit moves against you when memory effects develop. I see CC as really useful only to get a measure of how much more is potentially left to recharge when you realise that you are getting terminated early.
Termination based on CC in general is just a recipe for overcharging because CC is highly unreliable unless recalibration has just occurred. When you are getting interrupted charge cycles, CC is a random number not worth having unless you can suddenly reset it at the bottom. A BMS could do that, but not your average add-on gadget for LAs.

Anyway, this is broadly what I have been thinking of implementing to try and clear the cell memory. Because it took years to get into the interesting situation I am in, I want to proceed in a very controlled and documented way.

[tanglewood]

Originally Posted by john61ct

It is a real thing, but not a significant problem if you are aware of it and proactively take steps to prevent it.

Thousands of cycles without any capacity drop below the mfg Ah rating is not difficult to achieve, nor is the necessary gear that expensive.

But you do need to **care** about properly looking after your bank and go to the trouble to learn a fair bit, especially if you're trying to DIY from bare cells, own parts sourcing and system design.

If you reckon, a decade's goof enough, I'll just replace when needed anyway, then no worries do that.

So you've experienced it, and know how to mitigate?

[evm1024]

Just a few more datapoints - I charged the bank and got all cells over 3.5 VPC. There is a spread, the 'weaker" cells voltages rise more quickly. Anyway I charged at about 50 amps (0.07C) until all cells were over 3.50 volts then let the cells rest for 2 days without load. Here are the resulting rest voltages.

cell 4) 3.338 v
cell 3) 3.335 v
cell 2) 3.334 v
cell 1) 3.338 v

Total 13.349 v

All measured with my Fluke 87V

There is a 4 mV difference between cell 4 (and 1) and 2 with cell 3 closer to the 2 cells at 3.338 v.

No balancing has been done in years.

Today I went ahead and charged each cell to 3.65 volts. This was done individually to a tail current of 28 amps (0.04C). Even so it took some time to get the cells to 3.650 volts. The better cells took much longer to come up 3.65 vpc (#1 and #4).

Right now they are resting and in a day or so I'll measure the rest voltage and then start a discharge cycle. To see what the capacity is and to see what vpc the "good" cells have when the not-so-good cells are at the cutoff voltage for the test.

Those with direct experience will note that I did it this way to avoid the very long time that would be required to get all 4 cells in parallel to 3.65 vpc.

The next step will be to charge that bank again to bring the cells up near 3.65 VPC, then break the pack apart and charge them the rest of the way in parallel. Then of course another discharge cycle.

Hmm, I could buy a 100+ amp 5 volt power supply from the local surplus place and reset the voltage to 3.65 volts.... Sometimes the go for $25.

[john61ct]

Great stuff, thanks!

Isn't that a much longer resting period than needed?

I bet if you do your best to restore capacity at these very low charge rates for quite a few cycles and results aren't what you'd like,

getting up to much higher say 1C if feasible may produce a very different result.

I've been picking these up for under $60 delivered

TDK SWS600L-3
nominal 3.3V , but 2.64 - 3.96V adjustable
NOT current limiting but OCP triggered at 126A or 105%, or 400W, and that's real, not like cheap kit ratings

Current cheapest one available:
https://www.ebay.com/ulk/itm/152921385769

_______
Not recommending over .3C for normal usage, just for this sort of test/maintenance protocol, occasional use only.

[OceanSeaSpray]

Great data. This kind of resting OCV after charging suggests cells in the 75-80% SOC range only, but note that you were definitely not fully absorbed with 0.07C at the end.

0.04C residual current is still above the typical C/30 termination point at 3.65V. Did you meter the additional charge going into the cells when top balancing again?

I also have signs of imbalance in my pack when charging, but not in the resting voltages afterwards. Since I can't reach the top and I hit a fake upper knee, I am holding off on any statements about cell imbalance at the moment. It is not the cause for the missing capacity in any case.

[evm1024]

Originally Posted by OceanSeaSpray

Great data. This kind of resting OCV after charging suggests cells in the 75-80% SOC range only, but note that you were definitely not fully absorbed with 0.07C at the end.

0.04C residual current is still above the typical C/30 termination point at 3.65V. Did you meter the additional charge going into the cells when top balancing again?

I also have signs of imbalance in my pack when charging, but not in the resting voltages afterwards. Since I can't reach the top and I hit a fake upper knee, I am holding off on any statements about cell imbalance at the moment. It is not the cause for the missing capacity in any case.

Yes, I do not consider what I did as a top balance. Just a quick and dirty test to bring them closer to top balance.

After the discharge (capacity test) I'll reprogram my Victron to higher specs and charge to near 3.65 vpc. Then break the pack and go for the balance.

As it sits the voltage hangs in at 3.65 vpc and climbing at 30 amps but falls to a (much) lower vpc at 25 amps. That tells me that we are a long way from "parity" at 3.65 vpc.

It will be interesting to see what the rest voltage becomes after sitting a day or few in the boat.

It would be great if they were actually only charged to 70-80% and could be "fully" charged.

We will see what this capacity test gives us and what the results of a "real" top balance to 3.65 vpc at low amps gives us.

[evm1024]

Originally Posted by OceanSeaSpray

Snip

0.04C residual current is still above the typical C/30 termination point at 3.65V. Did you meter the additional charge going into the cells when top balancing again?

I also have signs of imbalance in my pack when charging, but not in the resting voltages afterwards.
SNIP

I am looking forward to any data that you post. I'm doing the quick and dirty and I think I recall that you were going to be very methodical.

For charging I brought the whole pack up with the Victron inverter/charger then took my bench supply and set the voltage such that the supply went into current limit (at 30 amps). Once the voltage hit 3.65 at the cells I set the current limit to 28 amps and kept the open circuit voltage to 4.4 volts.

The supply was current limited (at 28 amps) and I monitored the cell voltage with my Fluke. When it hit 3.65 volts I switched to the next cell.

After with the cell in parallel the c/30 rate will be huge (93 amps!). I'll shoot for much less or do them individually again but use voltage as the setpoint.

I also don't think that imbalance is the problem. with a resting voltage within 4 mV they sound very balanced to me. As I recall under a moderate load they tended to be close to each other.

[OceanSeaSpray]

Originally Posted by evm1024

As it sits the voltage hangs in at 3.65 vpc and climbing at 30 amps but falls to a (much) lower vpc at 25 amps. That tells me that we are a long way from "parity" at 3.65 vpc.

That strong dependency between the cell voltage and current is a sign of high internal resistance to charging. This is exactly what I have also observed.

Originally Posted by evm1024

It will be interesting to see what the rest voltage becomes after sitting a day or few in the boat.

It would be great if they were actually only charged to 70-80% and could be "fully" charged.

Keep in mind that the longer you rest it, the more the hysteresis between the OCV after charging and OCV after discharging averages out. I would try to measure it after approximately 4 hours. It is normally long enough to get it settled.

The SOC is a little murky because of the 2-day rest, but I think you will find capacity upwards.

The simplest procedure might be not breaking the pack and just recording the charge going in as you keep absorbing. If there are balance problems they will become apparent at some point.

This is really interesting.

[evm1024]

A gentle reminder that this thread is for posting the observations and actions of folks who are experiencing (what may be) memory effects in their banks.

If you have a problem with a specific post please do report it to the mods.

If you have an outstanding idea but do not have a LiFePO4 bank please PM the poster that you want to respond to. Or post it in the more general LFP thread.

Please do follow the mods suggestion of ignoring posts rather than responding to. This will help keep the S/N ratio higher.

Hopefully we can keep this thread on track.

[newhaul]

Originally Posted by evm1024

A gentle reminder that this thread is for posting the observations and actions of folks who are experiencing (what may be) memory effects in their banks.

If you have a problem with a specific post please do report it to the mods.

If you have an outstanding idea but do not have a LiFePO4 bank please PM the poster that you want to respond to. Or post it in the more general LFP thread.

Please do follow the mods suggestion of ignoring posts rather than responding to. This will help keep the S/N ratio higher.

Hopefully we can keep this thread on track.

thank you for starting this thread .
As I am only 2 years in on my Lfp bank hopefully the knowledge I learn here will allow me to avoid this " memory" pitfall going forward.

[evm1024]

I got down to the boat again this morning and measured each cells voltage. This represents a 11 hour rest period.

cell 4) 3.343
cell 3) 3.341
cell 2) 3.341
cell 1) 3.343

Cells 2 and 3 were bouncing back and forth between 3.341 and 3.342 volts so I just rounded (down) to the greater spread (3.341). This shows that the bank has a 2 mV spread and "should" be top balanced.

I've reconnected the bank to the boats systems and started the discharge cycle. I only have enough incandescent light bulbs (120 volt) to get a 10 amp draw from the bank. I'll have to get a few more to get up to 25 or 30 amp draw.

As you might infer I am depending on the inverter/charger to shutoff the draw at the low end of the curve.

[OceanSeaSpray]

Originally Posted by evm1024

I got down to the boat again this morning and measured each cells voltage. This represents a 11 hour rest period.

cell 4) 3.343
cell 3) 3.341
cell 2) 3.341
cell 1) 3.343

Cells 2 and 3 were bouncing back and forth between 3.341 and 3.342 volts so I just rounded (down) to the greater spread (3.341). This shows that the bank has a 2 mV spread and "should" be top balanced.

The OCV curve I am using at has a datapoint for 78% SOC at 3.344V following a charge. It is consistent with the previous figures you posted, maybe just slightly higher.
When I attempted a second charge cycle after absorbing the cells to 3.5V and C/30, I also found that I had made almost no headway in terms of reclaiming capacity. Any attempt at increasing the current even slightly sent the cell voltages up - high resistance.

Same situation, since the cells are not charged enough, balance issues might be hidden. Balance issues are not normally visible at 3.34V.

Originally Posted by evm1024

I've reconnected the bank to the boats systems and started the discharge cycle. I only have enough incandescent light bulbs (120 volt) to get a 10 amp draw from the bank. I'll have to get a few more to get up to 25 or 30 amp draw.

As you might infer I am depending on the inverter/charger to shutoff the draw at the low end of the curve.

The voltage falls off so quickly at the bottom end, if you get the cells to rest around or below 3.0V once discharged, they should be within 1-2% of empty. I discharged naturally through normal consumption until the BMS warning alarm was triggered, so I am in the same situation basically.

My bet here would be that you will recharge back to the same point and hit the wall again, but try to log the charge going in as much as possible, so you have an idea of the upside that should exist.
If that happens, you won't have too many other options but keep putting current in and the only variables left will be 1/ at what voltage and 2/ how long for/how much capacity.

[Delfin]

Originally Posted by newhaul

thank you for starting this thread .
As I am only 2 years in on my Lfp bank hopefully the knowledge I learn here will allow me to avoid this " memory" pitfall going forward.

Yes, this is a very useful thread. My bank is also 2 years in, and after reading all opinions have come to the following conclusions on how to manage my LFP bank:

1. As MaineSail has noted, hitting voltages of 3.65 per cell isn't what damages the cells, it is the duration at any voltage above around 3.35 vpc. The advice of some is to avoid the "knees" for longevity, but this probably comes at a cost of diminished total capacity as the result of continual undercharging and the dreaded memory effect. Consequently, a full recharge every 25 cycles or so to 3.65 volts with a 30 minute absorption will minimize memory effects.

2. I suspect, but don't know, that it may also be helpful to discharge the batteries periodically to 15% or so SoC. I am going to try to do this in conjunction with the full recharge to 3.65 volts per cell. In other words, exercise the batteries, which is consistent with the mfg. recommendations.

I have noted that with continued charge termination at 3.55 vpc, the resting voltage has crept down minutely over time. By resting voltage, I mean about an hour after charge termination with no load. On reflection, this supports the notion that a gradual decrease in capacity due to continued PSOC has occurred. I'll know more after a few more full charges to 100% SoC at 3.65 vpc and absorption amps of 1 - 2% for 30 minutes, but after the first and only full charge under this regime I've given this bank I noted a resting voltage .2 volts higher over than previously.

FYI, the bank is two Lithionics 8s 300 Ah 24vdc packs.