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

[NahanniV]

Originally Posted by mcarling

The problem is that the current counter will drift away from an accurate reading over a number of charge/discharge cycles -- unless supplemented by another means of knowing the SoC (probably near the extremes) to self-calibrate when possible.

Yes, I agree, the SOC will get out of sync and need to be reset; That reset will likely also need to sense current to work properly. Even my old Link-20 handles this quite well.

[Lagoon4us]

Originally Posted by NahanniV

State of charge can only be done well by counting current, it's the only way that gives useful results in a working system; This was also the case for Lead Acid.

I wish there was an integrated system that did Cell level battery monitoring of temp and voltage with alarms and a state of charge display (at a reasonable price).

The Elite Power Solutions BMS: http://elitepowersolutions.com/docs/EMS.pdf
looks good but only works easily for GBS cells.

I've been playing with the PowerLog 6s. For $40 it can do a lot: cell monitoring, 3 temp inputs, 1 alarm output, current measure 40A continuous and 140 peak, current counting, but no simple state of charge display and the current limit is too small for all loads (start, windless). I may Mod it to an external current shunt.

Anybody Know of another system ?

I need to "put a stake in the ground" and start planning an install soon. Only 70 days till launch !

Of any use?
http://www.ev-power.eu/docs/pdf/BMS123_Manual_v1_0.pdf

[charlie p]

Yes, there is a cumulative error problem with amp/hour meters. Its been around for decades and the vendors have tuned their software to minimize the error. They are still quite useful - I know I wouldn't give mine up.

As far as LFP batteries go, it seems to me that cumulative meters might actually work a little better than with LA. The sharp knees at both ends of the charge/discharge curve give precise recalibration points. Also, to get good calibration, you need to hit both ends. LA banks don't like to be fully discharged, so calibration tends to be incomplete. With LFP, there is no real harm to full discharge, so more accurate calibration is possible. In fact, since a major advantage of LFP is the ability to use more of the capacity range, it is likely that low SOC cal will be more common. If getting down to about 20% is routine, then occasionally dropping down to zero isn't a big deal. No need to burn a full charge just to get a cal data point.

At the high SOC end, LFP could be better because it can take high charging currents until close (though not quite at) full SOC. With LA, cruisers who rely on engine charging often just cycle between 50% and about 85%. Running the engine to get the last 15% at low current just isn't worth it. With LFP, engine charging can get to full SOC, thus providing frequent calibration points.

Given that cumulative meters are useful (though imperfect) today with LA and are likely to be better with LFP, I would say there is not much to worry about. Is there anyone here who has been using LFP with a cumulative meter who could provide some hard data?

Another thing to think about. Since LFP is being driven, so to speak, by the electric car industry, and it is obsessed with SOC measurement, it is likely that better measurement technology will emerge.

Charlie

[ebaugh]

Originally Posted by charlie p

Yes, there is a cumulative error problem with amp/hour meters. Its been around for decades and the vendors have tuned their software to minimize the error. They are still quite useful - I know I wouldn't give mine up.

As far as LFP batteries go, it seems to me that cumulative meters might actually work a little better than with LA. The sharp knees at both ends of the charge/discharge curve give precise recalibration points. Also, to get good calibration, you need to hit both ends. LA banks don't like to be fully discharged, so calibration tends to be incomplete. With LFP, there is no real harm to full discharge, so more accurate calibration is possible. In fact, since a major advantage of LFP is the ability to use more of the capacity range, it is likely that low SOC cal will be more common. If getting down to about 20% is routine, then occasionally dropping down to zero isn't a big deal. No need to burn a full charge just to get a cal data point.

At the high SOC end, LFP could be better because it can take high charging currents until close (though not quite at) full SOC. With LA, cruisers who rely on engine charging often just cycle between 50% and about 85%. Running the engine to get the last 15% at low current just isn't worth it. With LFP, engine charging can get to full SOC, thus providing frequent calibration points.

Given that cumulative meters are useful (though imperfect) today with LA and are likely to be better with LFP, I would say there is not much to worry about. Is there anyone here who has been using LFP with a cumulative meter who could provide some hard data?

Another thing to think about. Since LFP is being driven, so to speak, by the electric car industry, and it is obsessed with SOC measurement, it is likely that better measurement technology will emerge.

Charlie

I have the Elite EMS. It has a SOC counter that seems to keep pretty close, but not perfect unless it resets on the high end. No good measurements though since I can't take the pack offline to do any real checking. But I've easily done 30 sequential partial discharges in the 3.2 to 3.4V per cell range represented as 40 to 90% SOC and it's never given silly readings. The EV guys almost always go back to 100% every charge. So they have it easy.

[diugo]

Current counting requires an accurate Peukert value, otherwise the counter gets out of sync quickly and forces either a full charge or discharge to recalibrate. But, it's simpler to implement and, as pointed out, there are existing devices out there that already do this.

But a meter that determines SoC based simply on voltage and current won't need to be recalibrated nearly as often and can avoid the knee regions indefinitely. The problem is that the data needs to be gleaned---either by loading an external database or by self-learning.

It only takes two steady full discharges to calculate the Peukert exponent; interpolation may be good enough for currents in between. It's a PITA that a low-current learn pretty much takes forever!

[Lagoon4us]

Of interest for a monolithic battery assembly!

The original Winston LP12V80AH battery - after 3 years at 98AH!
The photo shows the LP12V 80AH battery produced by Winston (ThunderSky) in 2010. After 3 years of continueous use it gives 98 AH (!) of real capacity. After the warranty period, the battery was modified by the user to get access to each cell and to monitor the voltage of the cells, using the CellLog8.
What can be said? It seems better than “ideal dream”. But it is not a dream, it is the reality of the LFP technology. With LFP cells energy storage is possible!

[goboatingnow]

Originally Posted by diugo

Current counting requires an accurate Peukert value, otherwise the counter gets out of sync quickly and forces either a full charge or discharge to recalibrate. But, it's simpler to implement and, as pointed out, there are existing devices out there that already do this.

But a meter that determines SoC based simply on voltage and current won't need to be recalibrated nearly as often and can avoid the knee regions indefinitely. The problem is that the data needs to be gleaned---either by loading an external database or by self-learning.

It only takes two steady full discharges to calculate the Peukert exponent; interpolation may be good enough for currents in between. It's a PITA that a low-current learn pretty much takes forever!

My experience is mainly with lower AH cylindrical cells , where the linear region has a defined slope and voltage is almost always used as an SOC.

The problem with high Ah prismatics is that you can't just use amp counting you have to have a reset , only one is needed usually on full charge. This typically uses something like a 2-5% of C in amps as a reset to full

Unfortunately this is difficult in LFP, you get very close to the knee. In practice for long life you never want to bring these cells to100%. Hence the difficulty in keeping these current counters devices in sync.

Dave

[nimblemotors]

Originally Posted by charlie p

Yes, there is a cumulative error problem with amp/hour meters. Its been around for decades and the vendors have tuned their software to minimize the error. They are still quite useful - I know I wouldn't give mine up.

As far as LFP batteries go, it seems to me that cumulative meters might actually work a little better than with LA. The sharp knees at both ends of the charge/discharge curve give precise recalibration points. Also, to get good calibration, you need to hit both ends. LA banks don't like to be fully discharged, so calibration tends to be incomplete. With LFP, there is no real harm to full discharge, so more accurate calibration is possible. In fact, since a major advantage of LFP is the ability to use more of the capacity range, it is likely that low SOC cal will be more common. If getting down to about 20% is routine, then occasionally dropping down to zero isn't a big deal. No need to burn a full charge just to get a cal data point.

At the high SOC end, LFP could be better because it can take high charging currents until close (though not quite at) full SOC. With LA, cruisers who rely on engine charging often just cycle between 50% and about 85%. Running the engine to get the last 15% at low current just isn't worth it. With LFP, engine charging can get to full SOC, thus providing frequent calibration points.

Given that cumulative meters are useful (though imperfect) today with LA and are likely to be better with LFP, I would say there is not much to worry about. Is there anyone here who has been using LFP with a cumulative meter who could provide some hard data?

Another thing to think about. Since LFP is being driven, so to speak, by the electric car industry, and it is obsessed with SOC measurement, it is likely that better measurement technology will emerge.

Charlie

I have no hard data, but you've got it right, by monitoring the voltages a system can determine when its amp counting needs to reset. It should not occur very often unless something is going wrong, once the measured capacity declines to a certain level, you raise an alarm.
The Prius does this also.

JackB

[goboatingnow]

Originally Posted by nimblemotors

I have no hard data, but you've got it right, by monitoring the voltages a system can determine when its amp counting needs to reset. It should not occur very often unless something is going wrong, once the measured capacity declines to a certain level, you raise an alarm.
The Prius does this also.

JackB

The method outlined in the previous post is not practical in boats. ( or is difficult ). Firstly with Li ferrous we generally stay away from the knee , in reality the cells perform best at about 90% SOC. at that SOC there is still significant charging current flowing, hence the reset on full algorithm tends to difficult to ascertain. ( most counters use a voltage set point and combine it with a low current set point

Running the battery down to close to 0 % simply to reset a current counter is of course madness in real life. , firstly you get very close to the LVC point and in most case it will trigger first.

So this leaves is in the tricky position in real of possibly never getting a really full or empty reset. Amp counters are notoriously error prone where such conditions occur and in practice you have no way of cross checking them.

Yes they can work and I've designed some into my software at times , but they have significant issues in Li environments.

Dave

[diugo]

I think you will find there are two boat owner camps. The first doesn't mind charging the pack to nearly 100% every time---that way, they know the battery is fully charged and will provide the longest run time if necessary. An automatic amp-hour reset at the top will provide an SoC that is close enough, especially if the correct Peukert value is used.

The other camp wants their expensive batteries to last as long as possible---up to the rumored ten years---and will therefore keep a constant battery eye, ensuring the pack stays in the 20-80% longevity zone as much as possible. They will need to more accurately infer SoC from the voltage and current.

[goboatingnow]

Originally Posted by diugo

I think you will find there are two boat owner camps. The first doesn't mind charging the pack to nearly 100% every time---that way, they know the battery is fully charged and will provide the longest run time if necessary. An automatic amp-hour reset at the top will provide an SoC that is close enough, especially if the correct Peukert value is used.

The other camp wants their expensive batteries to last as long as possible---up to the rumored ten years---and will therefore keep a constant battery eye, ensuring the pack stays in the 20-80% longevity zone as much as possible. They will need to more accurately infer SoC from the voltage and current.

I don't agree 100% SOC is neither desirable nor easy to control with typical boat systems. Float or trickle charging is very bad for Li. If you read back through the posts you will see a general caution to suspend charging at well below the 100% point ( outside of lab tests)

If you look at the charge profile of Li its very hard using current to determine a reset point, the cells accept current at high rates upto and past the point of overcharge. So all we can do is pick a voltage /current combination and reset from there. Peukerts has no real applicability in Li. Values are close to unity and what I see can vary a lot depending on high C discharge environments or low C ones. The big issue is that capacity of the cells is hard to determine , hence comparing current counting as a measure of accuracy is error prone coupled with losses in recharging etc.

Amp counting is ok for " fuel gauge " type scenarios , I certainly would not rely on it to control chargers or cutoff devices. Voltage ( sometimes with current ) is really the only way for that. Other then that amp counting is a good. Way to give some feedback as to what's left in the batteries

Dave

[diugo]

I keep harping on Peukert b/c it seems to go in one ear and out the other with a lot of the crowd here. Like it or not, Peukert is nowhere near unity (ranging from 1.04 in my case to 1.12 for MaineSail and 1.20 for DeckOfficer) and needs to be taken into account especially if accurate amp-hour counting is desired.

I agree with you that 100% SoC charging should be avoided for longevity. However, in practice, the limit will end up being pushed by many, mostly on account that a certain highly revered "someone" here keeps insisting that 100% is perfectly okay---since battery current is "zero" when fully charged.

[goboatingnow]

Originally Posted by diugo

I keep harping on Peukert b/c it seems to go in one ear and out the other with a lot of the crowd here. Like it or not, Peukert is nowhere near unity (ranging from 1.04 in my case to 1.12 for MaineSail and 1.20 for DeckOfficer) and needs to be taken into account especially if accurate amp-hour counting is desired.

I agree with you that 100% SoC charging should be avoided for longevity. However, in practice, the limit will end up being pushed by many, mostly on account that a certain highly revered "someone" here keeps insisting that 100% is perfectly okay---since current is "zero" there.

It's worth pointing out that Peukerts Law is a model of LA discharge . It is not a model of Li discharge. It can approximate, but that's all. In reality you cannot compare the two systems.

100 % SOC cannot be determined using current. That's why we use CC-CV charging. In effect you pick a voltage set point and if that set point is low enough the current will not flow into the battery. What the SOC is can only be estimated by subsequent discharges. Anyone seriously seeking 100% charging of large capacity prismatics need to know what they are doing , and it does the cells no good.

Dave

[Lagoon4us]

So the MPPT i have will not go higher than 14.8 volts unless a fault occurs, maintenance rate is 14.1 all good.
If the cell logger detects the previously mentioned fault, the alarm sounds and disconnect occurs.
The amp counting/fuel gauge/monitor is only for our convenience i thought. At less than 3.70 charge rate per cell on the highest selection on the MPPT surely we are all safe?

[diugo]

Originally Posted by goboatingnow

It's worth pointing out that Peukerts Law is a model of LA discharge . It is not a model of Li discharge. It can approximate, but that's all. In reality you cannot compare the two systems.

That is totally incorrect. Peukert has been shown to apply to all battery chemistries. It applies perfectly to ZiC, NiFe, NiCd, NiMH, other Li-ion chemistries, and yes---even to vaulted LFPs. LiFePO4 is very good, but still far from perfect.