How do you charge your LiFePo4 Batterybanks?

[hjohnson]

Originally Posted by fxykty

I’m very happy that you’re happy, but you’re going to kill your LFP battery treating it like a LA battery.

Why are you still charging the battery when it is basically at 100% SOC to hold it there? LFP is not LA - it is not good (in fact, quite bad in terms of retaining capacity) to charge and hold an LFP battery anywhere above ~50%.

Once you charge your battery to 100%, TURN OFF ALL CHARGING. Let the battery discharge to below 30%, then turn on charging again.

To repeat, LFP are not LA and the LA habit of getting to and holding at 100% is anathema for LFP. LFP likes to be held at 50% and cycle through 20-80%.

https://nordkyndesign.com/charging-m...battery-banks/

BTW, 14.5V charging and 10A trailing is pretty high - 13.9V and 2.5% of C trailing amps is more appropriate as a conservative 100% SOC.

From the research that I’ve done, the wear on LiFePO4 occurs when holding the battery up at the knees of its charge curve (ie the 3.65v/cell). That’s explicitly not what I’m doing. It hits that when on a charge cycle extremely briefly, then quickly drops down to 13.9V. it takes a few minutes to get there, with all the charging sources disabled, then it ramps things back up to hold 13.85V or so, with no current at all flowing in or out of the battery.

[Kit_L]

Very interesting thread; joining simply to learn more about a technology I intend to embrace soon.

[CaptainRivet]

Originally Posted by fxykty

Also, for best longevity of an LFP battery you certainly don’t want to be regularly charging to 100% more than once a month or so. And ensure that whenever you charge above 80% that you have a reasonable load applied so that the battery doesn’t rest at that level.

Nice theory but in reality, especially with typically bigger banks of bluewatercruisers they are cycling always different but majority its 60-100%SOC, that looks like the following:
Well your BMS normally starts a new charge cycle every day means it will charge till the parameter set for end of charge, typically for marine banks somewhere between 3,45 and 3,55V which is 95-98% full and the 100%SOC reference for the BMS. Don't know any BMS where you can define the length of the charge cycle to eg every 3 or 5dayes instead daily or set/define charge every 30 days charge to 3,65V.
Requested that as software change on my electrodacus BMS and it got denied.
It works typically as follow:
So your bank will hit the 100% SOK on a good sunny day at around noon till 2pm. Then you dump the excess solar power into hot water heater and/or run the watermaker and/or Washing machine. Some lunch burns of some if you have an electric galley. Sundowner icemaker but still 100% SOK. Then dinner in best case taken from bank, typically standby overnight from bank SOK 70-80% and a new charge cycle starts. Day cloudy so bank is only full at sundowner and next day cycle starts at 60% SOK,another cloudy day and you are at 45-50SOK...

[fxykty]

Originally Posted by CaptainRivet

…

Well your BMS normally starts a new charge cycle every day means it will charge till the parameter set for end of charge, typically for marine banks somewhere between 3,45 and 3,55V which is 95-98% full and the 100%SOC reference for the BMS. Don't know any BMS where you can define the length of the charge cycle to eg every 3 or 5dayes instead daily or set/define charge every 30 days charge to 3,65V.

Requested that as software change on my electrodacus BMS and it got denied.

It works typically as follow:

...

So get a better BMS that can enforce charge profiles that use SOC rather than voltage. The TAO BMS does exactly what you requested of Electrodacus: https://www.taoperf.com/2020/12/10/n...le-management/

[fxykty]

Originally Posted by hjohnson

From the research that I’ve done, the wear on LiFePO4 occurs when holding the battery up at the knees of its charge curve (ie the 3.65v/cell). That’s explicitly not what I’m doing. It hits that when on a charge cycle extremely briefly, then quickly drops down to 13.9V. it takes a few minutes to get there, with all the charging sources disabled, then it ramps things back up to hold 13.85V or so, with no current at all flowing in or out of the battery.

13.85V is 3.46V per cell. That’s way too high to hold an LFP battery at - about 90% SOC. Didn’t you read the reference I provided for why that’s not a good idea if you want to preserve your LFP battery’s capacity?

[CaptainRivet]

Originally Posted by fxykty

So get a better BMS that can enforce charge profiles that use SOC rather than voltage. The TAO BMS does exactly what you requested of Electrodacus: https://www.taoperf.com/2020/12/10/n...le-management/

So Phil Tao implemented that now. Great BMS, only one that has that.
Electrodacus is great too for 1/10 of the price does everything what you want, except of this feature. You can trick Electrodacus by just setting the charge paramters accordingly, you can chose between voltage and SoC. This means end of charge parmeter at 80% and charge recovery at 30% will result in the same. I only need to change manually all 30 days end of charge to 100%soc and after full change back.

[hjohnson]

Originally Posted by fxykty

13.85V is 3.46V per cell. That’s way too high to hold an LFP battery at - about 90% SOC. Didn’t you read the reference I provided for why that’s not a good idea if you want to preserve your LFP battery’s capacity?

Yes, and your reference is just one of many out on the internet. It’s not gospel.

[goboatingnow]

How are bmses interpreting soc without using voltage , coulomb counting is very poor at providing reliable soc

[Philtao]

Originally Posted by goboatingnow

How are bmses interpreting soc without using voltage , coulomb counting is very poor at providing reliable soc

You are right:

• for a 500A/50mV shunt, a variation of current of 0.1 A corresponds to a shunt voltage variation of 10 µV
• if the current measure is wrong by 0.1 A, that accumulates to 72 Ah over a month.

So if you need a precise instrument capable of measuring micro-volts (or 10 mA) in order to have a reliable SOC estimate over time.

For the BMS, I use a high precision instrumentation amplifier with a programmable gain in order to have accurate current measure when current is low (which is most of the time). Then the algorithm is self learning to adjust some parameters every time the SOC is re-calibrated to 100%.

Result: I do a full charge every month so that the BMS re-calibrates the SOC to 100%. My experience is that when the SOC resets to 100% (based on voltage and tail current) the error of the computed SOC is less than 2.5%. The following picture shows the situation when the battery was last full (May 1st @ 12:44:33). The step on the SOC graph corresponds to the accumulated SOC error since the last full charge that was on March 29th (SOC error = 2.1% after nearly 5 weeks)

I am not sure a voltage based SOC estimation can be that precise (?)

[CharlieJ]

@GBN #38

How are bmses interpreting soc without using voltage , coulomb counting is very poor at providing reliable soc

This is incorrect for LFP batteries. The charge/discharge curve is so flat that a few tenths of a volt can equate to 10% difference in SOC. Voltage is essentially useless for determining SOC of LFP batteries. But I think you know this.

[goboatingnow]

Originally Posted by CharlieJ

@GBN #38

This is incorrect for LFP batteries. The charge/discharge curve is so flat that a few tenths of a volt can equate to 10% difference in SOC. Voltage is essentially useless for determining SOC of LFP batteries. But I think you know this.

Yes but all the ICs on the market use voltage

The point with LFP is that the voltage soc curve is not flat and is linear. This means soc estimation by accurate voltage is reasonable, especially if your software allows battery calibration curves to be stored.

Since with LFP , you have to use voltage endpoints to control charge , you can’t simply use coulomb counting , and, in general knowing your battery is at 65.4% SOC is a bit esoteric.

I use voltage on my bms and I find that the results are typically within 10-15% of coulomb counting ( this is with calibrated voltage curves )

Hence I see little utility in coulomb counting

In LA it’s useful because the V/SOC curve is complex and highly load dependant.

[CaptainRivet]

Originally Posted by goboatingnow

Yes but all the ICs on the market use voltage

The point with LFP is that the voltage soc curve is not flat and is linear. This means soc estimation by accurate voltage is reasonable, especially if your software allows battery calibration curves to be stored.

Since with LFP , you have to use voltage endpoints to control charge , you can’t simply use coulomb counting , and, in general knowing your battery is at 65.4% SOC is a bit esoteric.

I use voltage on my bms and I find that the results are typically within 10-15% of coulomb counting ( this is with calibrated voltage curves )

Hence I see little utility in coulomb counting

In LA it’s useful because the V/SOC curve is complex and highly load dependant.

Well that it kinda wrong, very good and advanced BMS like REC, Tao and Electrodacus use minimum a combination of voltage, coulomb counting and current. Like this you can identify quite precise at which SoC yo are, especially in the upper and lower knee parts and the self learing algorithm narrows/calibrates that more and more down but on the other hand as Phil Tao explained the measuring error counts up. In the flat curve part in between the 2 knees additionally counting whats going in/out additionally narrows down your SOC precision, thats why electrodacus has 2 shunts (one load, one charge) in the plus pole and that works surprisingly well. The load also sees the charge side and can calibrate with this the shunts quite well too.
And because of the meassuring errors count up you have to give it a full charge so the algorithm can calibrate/adjust itself.

[goboatingnow]

Originally Posted by CaptainRivet

Well that it kinda wrong, very good and advanced BMS like REC, Tao and Electrodacus use minimum a combination of voltage, coulomb counting and current. Like this you can identify quite precise at which SoC yo are, especially in the upper and lower knee parts and the self learing algorithm narrows/calibrates that more and more down but on the other hand as Phil Tao explained the measuring error counts up. In the flat curve part in between the 2 knees additionally counting whats going in/out additionally narrows down your SOC precision, thats why electrodacus has 2 shunts (one load, one charge) in the plus pole and that works surprisingly well. The load also sees the charge side and can calibrate with this the shunts quite well too.
And because of the meassuring errors count up you have to give it a full charge so the algorithm can calibrate/adjust itself.

Yep shunts have nothing to do with the question at hand any decent system must be able to distinguish between charge current and tail current.

The point remains large scale semiconductor
Manufacturers with state of the art BMS chips sets don’t rely on coulomb counting.

I’m not discounting coulomb counting merely for LFP you cannot make protection decisions on SOC. voltage is the key to protecting lithium’s

( ps cause a guy writes long and descriptive blogs arguing certain cases doesn’t necessarily make the product any better )

[mitiempo]

Originally Posted by CaptainRivet

...... that's why electrodacus has 2 shunts (one load, one charge) in the plus pole and that works surprisingly well. The load also sees the charge side.....

Yes Electrodacus has 2 shunts - the main one measures everything in and out. The other measures solar in only.

[CaptainRivet]

Originally Posted by mitiempo

Yes Electrodacus has 2 shunts - the main one measures everything in and out. The other measures solar in only.

Well if you only put solar on the PV shunt it only shows solar
Electrodacus was developed for offgrid shed with his DSSR controller and solar panels matching the battery voltage so the only charge source there is solar.
If you put it in an RV or boat with multiple charge sources they should go on the PV shunt too. Like this the current in/out measure is more precise and therefore SOC too. Additionally you have also a clean curcuit design with load and charge bus.
I have it this way confirmed by Dacian.
Just nobody asked so far and explained it to him why this is different to his shed setup. Dacian has his way of seeing things. He is still optimising it with clear focus on offgrid shed hence not wanting to implement the variable 100% soc charge instead daily. His bank is nearly 80% used daily and cycled hard, so daily full charge maximize his capacity. Li is expensive and solar+room for it is cheap... Different to boat where you typically have 3-5days capacity in the bank and very limited space for solar so you go for 50V bifacial panels+Mppt to optimize harvest from the limited space available.