LiFePO4 Batteries: Discussion Thread for Those Using Them as House Banks

[T1 Terry]

Originally Posted by T1 Terry

There are few different combinations and 2 different battery packs but I'll try and explain the ideas behind each lot of testing
This is the first pack, 360Ah @12v, made up of 90Ah cells, in total 3 boxes of them this size.
This set up is to give an example of size compared to 2 x 6V 125Ah Yuasa AGM on the left and a 105Ah Vision AGM on the right, 250Ah @12v of Yuasa AGM behind. The test was to see if connecting the cells and load across the end of the string would produce a difference in cell capacity across the pack after 3 mths of service, the answer is, no it didn't, all 16 cells had 90 Ah plus capacity at the end of the test.

That battery pack was reconfigured into a 180Ah @24v and a 90Ah @12v of the earlier LFP cells added on the end, this was for a motorhome setup. This set up has run 24/7 for the last 2 yrs, the owner lives in his motorhome on the road travelling. the second photo shows the slide out battery box in the side of the bus.

[deckofficer]

Originally Posted by T1 Terry

There are few different combinations and 2 different battery packs but I'll try and explain the ideas behind each lot of testing
This is the first pack, 360Ah @12v, made up of 90Ah cells, in total 3 boxes of them this size.
This set up is to give an example of size compared to 2 x 6V 125Ah Yuasa AGM on the left and a 105Ah Vision AGM on the right, 250Ah @12v of Yuasa AGM behind. The test was to see if connecting the cells and load across the end of the string would produce a difference in cell capacity across the pack after 3 mths of service, the answer is, no it didn't, all 16 cells had 90 Ah plus capacity at the end of the test.

Thanks Terry, now my memory is jogged. You have both the newer cells and older ones like mine. What threw me was 360 a-hr, I knew the common sizes, knew 360 a-hr wasn't one, so divided by 2 and that wasn't a size, divide by 3, that wasn't a size and I didn't think you were using the 90 a-hr, but I won't forget now.

Now that I see the installation for the first time, 90 a-hr makes perfect sense as you were able to max out your battery tray. Nice job.

Is the 24 volt bank for his inverter?

[T1 Terry]

Originally Posted by T1 Terry

That battery pack was reconfigured into a 180Ah @24v and a 90Ah @12v of the earlier LFP cells added on the end, this was for a motorhome setup. This set up has run 24/7 for the last 2 yrs, the owner lives in his motorhome on the road travelling. the second photo shows the slide out battery box in the side of the bus.

The next test was to build a 360Ah @ 12v set from half LYP cells and half LFP cells and see how they worked together, there were no issues. Next a made a duplicate set and tried running them in parllel as 2 x 360Ah batteries, they were so far out of balance with each other with a week one set sounded a low cell alarm in the early hrs of the morning yet the other set were still at 13.2v, they do not work well together as parallel batteries. I then connected them together in parallel at cell level. They stayed fairly close in balance like this, but one pack was always slightly behind the others. I don;t have photo of the present set up, each link strip is paralleled with cables and the link plates at alternate ends, forcing the charge/load across every cell, they have stayed in balance for 3 mths this way. A few days ago there was an unplanned capacity test, the sun didn't play the game properly and the inverters screamed for mercy at 10.5v, they were still pulling 35 amps even with the cells that low. The Ah meter reported 726Ah had been drawn, 726ah went back in before the cells all reached over 3.5v, I reset the 100% SOC and made a mental note to check what SOC the cells were in each night before going to bed.

[T1 Terry]

Originally Posted by deckofficer

Thanks Terry, now my memory is jogged. You have both the newer cells and older ones like mine. What threw me was 360 a-hr, I knew the common sizes, knew 360 a-hr wasn't one, so divided by 2 and that wasn't a size, divide by 3, that wasn't a size and I didn't think you were using the 90 a-hr, but I won't forget now.

Now that I see the installation for the first time, 90 a-hr makes perfect sense as you were able to max out your battery tray. Nice job.

Is the 24 volt bank for his inverter?

The24v pack runs his 2.5kW inverter, the whole bus is now 240V electric including, 250ltr fridge/freezer, cooking and hot water, he removed all the LPG stuff. There has since been a 180Ah @24v start pack added and these to are linked via a solenoid. The 12v set is recharged via 20 amp? 24v to 12v converter and powers stereos and GPS, UHF, modem, computer and the fridge in his tailer, the battery is really just a big capacitor to smooth the load.

T1 Terry

[deckofficer]

I might be using some of the smallest cells (40 a-hr) for my hot rod electric scooter being that I have a $280 credit from my last purchase from Balqon. Currently the scooter has been using Odyssey PC-680, (4) in series for a total of 720 Whr. At 48 volts the scooter is geared for 43 mph but the controller I'm using is good to 72 volts, 400 amps, so I might go with (24) 40 a-hr cells. Since it has 10" wheels I won't bother to go 65 mph the new voltage will achieve, but the range will increase from 6.5 miles per charge to 62 miles. For that added range the battery weight goes from 62 lbs to 79 lbs. I like that energy density over LA, 17 lbs increase for almost a 10 fold range increase.

[goboatingnow]

Originally Posted by mcarling

I don't think Charlie was suggesting that we choose a knee voltage. I think he was suggesting that we pick a threshold for dV/dSOC (the derivative of the voltage with respect to SOC i.e. the rate of change of the voltage as charging occurs i.e. the slope of the graph of Voltage against cumulative current). If so, I agree. I would use dV/dSOC rather than voltage as a threshold to identify the knees (just as we do when we look at charging graphs).

we dont have dv/dSOC measurements at the battery terminals , we have dv/dt ( and/or di/dt) thats all. We then assume a correlation to SOC. we may be right we may be wrong ( as the song goes)

In real life I fully agree with Terry , amp counting is a useful "fuel gauge" for general monitoring, its not a reliable way of determine SOC to any reasonable accuracy. Any system needs a deterministic and repeatable charge discharge cutoff and voltages are the only way in LI. ( and actually unlike LA, they are generally fixed and repeatable). What goes on between the two set points can to some extent only be guessed at.

Dave

[goboatingnow]

Originally Posted by daddle

At-rest volts is the accurate way to assess state of charge as a condition. That is what all the technical papers say. And that is what people are saying here even if they don't realize it. The problem is that the at rest condition is at least a few minutes for a rough reading, and according to some manufacturers up to 24 hours for accuracy. Hardly convenient. And, yes, the measurement in the middle region is significantly less accurate due to the flatness of the voltage/SoC function there. Then to determine the state of charge as Amp-hours one must perform a controlled calibrated charge-discharge to construct a table or function.

A huge amount of research has been done on complex ways to measure SoC while in use. The complex methods can be accurate, but they are unwieldy in practice.

.........

In practice the voltage curve of a LI cell is an accurate way to determine SOC, resting time for Li is in fact very short. The voltage curve during discharge is very linear ( the problem is in high C low discharge situations it can be virtually flat). Hence the majority of current LI-ion chargers uses voltage as a measure of SOC.( sometimes with current as a safety factor). I have several of them on my desk at the moment.

All I was trying to say is that amp counting is good for user "fuel gauges" its not sufficient to use for safety and charge cutoffs.

dave

[goboatingnow]

Originally Posted by nimblemotors

Huh? you see a "really full" when the voltage climbs, and "really empty" when it drops. You can detect the end points when the amp counting gets off.
i.e. you allow charge back in 10amps, but it goes "really full" at 8amps. There has been enough drift so you must reset. You now have 100% FULL, and can back off 10% from there. Same at the bottom. If you get too much reset, you must become more conservative on the counting.
Software has memory and can adapt, one of the great wonders of the 21st century.

I build all my own hardware and software, so I can't address what products do or don't do, in that sense I am of no help to anyone.

JackB

LI cells do not exhibit such current fall offs as you see in LA. The current ( depending on the cutoff voltage) may remain almost at max for the duration of the charge. Hence the issue of attempting fast charge near the 100% point. The battery simply never refuses current, unlike a LA which has a charge acceptance curve.

Terry has indicated his acceptable voltage range and I would agree, though with my custom setups on smaller batteries I can push a bit closer to the 100%. But with large capacity cells this is entering the "danger zone".

With Li, the battery doesn't indicate reliably a full position using current. Hence what you say cant really be applied. All you can do is to some extent arbitrarily select a cutoff point and use that voltage ( in reality you don't care at that point how much current is actually flowing into the batteries, the charge cycle terminates, end of story.

Dave

[goboatingnow]

It's kinda silly to say that since Ah meters are used, there is no Peukert effect.

Ohm's Law says a 4-millohm battery internally wastes less than a watt at 15 amps, but a whopping 90 watts at a charge current of 150A. Thus a single 150Ah charge cycle only returns 144Ah.

It's not about one small 6Ah difference---it is the 6Ah every charge cycle that quickly adds up and renders an uncorrected Ah meter useless, even as a crude guide.

I dont know your technical background, but you have a incorrect understanding of Peukerts law and its application.

Firslty Peukerts Law ONLY applies to the discharge of a battery, It does not model or attempt to model the charge cycle.

Secondly Peukerts Law, is only a "model" its not a direct battery effect. Peukerts was seeking a mathematical relationship to model and hence predict the loss of Wh capacity as a result of different discharge currents. Its merely a model and in real batteies can and do depart from the model,or in exotic chemistries like Li, have actual discharge profiles that differ from the Peukerts model. Ive yet to see an extensive study of the accuracy of model as it is applied to high capacity Li.

Note that the first order characteristic impedance of a battery is a complex value, it is comprised of static and dynamic elements (including real and reactive elements), simplistic applications can lead to errors of application. Higher order impedance models are quite complex and the subject of many Doctoral Thesis. ( even to this day for LA),

Dave

[Lagoon4us]

A good graph.

[mcarling]

Originally Posted by goboatingnow

we dont have dv/dSOC measurements at the battery terminals , we have dv/dt ( and/or di/dt) thats all.

An amp counting SoC meter obviously has access to its own SoC estimates and directly measures current flow in and out so, in practice, using dV/dSOC is not a problem.

[goboatingnow]

I need an accurate SoC "idiot gauge" for two reasons. Primarily, to ensure that my charger settings always keep my bank between 60-80%. I am hopeful that once the proper charge and float setpoints are methodically obtained, that the "60" figure can vary slightly according to my daily use, and still return to 80%.

Secondarily, if an extended cloudy period is forecast, I would likely want to bump the SoC to near 100% while the sun is still shining, to reduce the likelihood of running out of juice.


.......


I have cells. From Balqon, remember? I tried the simple straight amp-hour counting thing for a week, and got a lot more than a 1% error at charging currents not even exceeding 20A. That's why I'm on such a crusade, since readers here will likely experience much higher currents.

Think you need to read the online FAQ for your PL20 controller. It clearly says the displayed SoC can exceed 100%. More to the point, the PL series will automatically reset any SoC over 100% to 100% at the beginning of every discharge cycle. So your "1% error after 3 mths" was likely just 1% during the last cycle of your supposed test.

Amp counting as a way to make hard recharge cutoffs is a unobtainable quest. The method contains too many errors as the relationship between current and SOC is complex. Its useful as a rough fuel gauge and became popular in LA systems because of the complex relationship between terminal voltage and SOC in LA batteries ( though voltage model based meters are on the market , like Smart Gauge).

Amp hour counting especially in the mid range without a hard reset point, is doubly problematic and you have no independent way of verifying the accuracy of the displayed data ( until the lights go out).

Large capacity Li running at low discharge rates are doubly problematic and the voltage change in the linear range is very difficult to detect. SO all we really will know to any great extent is full ( an arbitrary set point picked by you) and "empty" another arbitrary set-point picked by the "user". In between theses points we will try an educated guesstimate.( the more characterisation you do the better the guess is).

LI technology should not really charged like LA, ie short recharges to stop the battery from going any way below 100%. This subjects Li to too many recharging cycles and this is the major cause of end of life failures.

The charging regime should really be

* charge to the determined upper setpoint, using CC followed by CV ( which is really all CV anyway)

* disconnect the charging source, do not float charge or undertake mini recharging cycles

* Re start charging either at the low set point or based on another arbitrary set point, you could use amp counting for this since its not a precise point, In my opinion you should minimise recharging cycles so leave the bank run down a fair bit. This of course is problematic for boats as the length and strength of the recharge cycle is often extremely variable and unknown.

* for long term storage ( like over wintering) the cells should be kept below 90% SOC and optionally between 40-60% SOC again amp counting is accurate enough to determine this.

Dave

[goboatingnow]

Originally Posted by mcarling

An amp counting SoC meter obviously has access to its own SoC estimates and directly measures current flow in and out so, in practice, using dV/dSOC is not a problem.

well it doesnt , what it has is a model of amps versus capacity, and applies in LA technology factors like charge efficiency ( an arbitrary number) and factors like Peukerts model to "attempt" to determine SOC. Its an educated guess.

All of them ( though not voltage models like Smart Gauge) require a "reset " point, a point where it is arbitrarily decided that the battery is full. In Li technology since we are doing the opposite to LA , in that we are trying to stay away from the 100% point, we have a greater difficulty in determining absolute SOC ( as opposed to relative SOC).

in LA batteries dv/SOC is a very difficult thing to measure , in Li its a little better but requires considerable accuracy in measurement and as yet we do not have a good handle of the variability of the dv/SOC for large prismatics, hence the time being spent to individually characterise them. I suspect we easily have a 10% or more error in production batteries

ALl Im saying is that amp counting to determine absolute SOC is very error prone and I have not seen any implementations that use it to make safety cutoff decisions. ( for example).

Dave

[Lagoon4us]

Small basic BMS units are available for under 500 bucks that give you a fuel gauge that resets, in the scheme of things worthwhile without overcomplicating the installation. It resets as the charger reaches full charge. If you look at the graph i posted in 1810# you set the reset point as you desire.

[mcarling]

Originally Posted by goboatingnow

well it doesnt , what it has is a model of amps versus capacity, and applies in LA technology factors like charge efficiency ( an arbitrary number) and factors like Peukerts model to "attempt" to determine SOC. Its an educated guess.

All of them ( though not voltage models like Smart Gauge) require a "reset " point, a point where it is arbitrarily decided that the battery is full. In Li technology since we are doing the opposite to LA , in that we are trying to stay away from the 100% point, we have a greater difficulty in determining absolute SOC ( as opposed to relative SOC).

in LA batteries dv/SOC is a very difficult thing to measure , in Li its a little better but requires considerable accuracy in measurement and as yet we do not have a good handle of the variability of the dv/SOC for large prismatics, hence the time being spent to individually characterise them. I suspect we easily have a 10% or more error in production batteries

ALl Im saying is that amp counting to determine absolute SOC is very error prone and I have not seen any implementations that use it to make safety cutoff decisions. ( for example).

I concede that using dV/dSOC to recalibrate is not trivially easy and that the resets might be less frequent than we might like. I have also not seen any implementations yet but, in principle, this is something that the SoC meter manufacturers can do when they decide the LiFePO4 market has become important enough and something that I expect they will do this year or next year. Is this method perfect? No, of course not. Is it good enough? I think so. Is there a better alternative? I don't see one.