Tesla battery pack?

[Sunking]

First of all, we don't measure storage systems in watts.

What are you talking about? Solar power systems are measured in Watts aka power. Energy or battery capacity is just an end result to meet daily energy requirements.

Not so sure you can call Golf Cart and EV Battery systems small as they are generally much larger than any off-grid or grid tied applications. How many off-grid and GT systems use 15 to 60 Kwh of storage? Answer very few and not the norm.

[Sunking]

Because counting always exhibits some drift, and because that error accumulates over time, that estimation technique can readily be off by more than 5%. It is only good for a short while, and then it is not so good any more. That is why it must be "resynchronized" regularly, and it is also why people are developing better estimators.

Agreed and that is my point, it drifts with every cycle. It is nothing more than the Hot Light or Check Engine Light.

To resync requires a full discharge capacity test consumers are not capable of doing. I have no problem with lithium batteries, they have their place and applications. They are not ready for consumer residential applications for the simple fact they are not cost effective and too dangerous. EV's you can justify the cost and dangers, but not in a residential application. As of today using lithium batteries in a residential applications are 400% more expensive, and run very high risk of thermal runaway destroying cells. When that reverses itself I will change my professional opinion. Until that time I will push back and counter so people can make an informed decision, not hype by advocates of one side.

[nebster]

At last count over 4,000,000 watts and counting in commercial and residential.

First of all, we don't measure storage systems in watts.

But, I strongly suspect that you don't own or manage a lithium chemistry-based residential or mobile ESS. It sounds to me like you've experimented with a few small-scale golf carts. I'd love for you to start a new thread, though, and tell us about your personal setup. That would be a lot more productive than acting like my seven-year-old in this one.

[jflorey2]

Here is the problem most folks do not understand and how you are nebster are spinning numbers. That 2S pack for the phone discharges at very low rates on the order of C72 to C/100 which for all practical purposes is Open Circuit Voltage.

Ha! During full power operation we'd be operating at 1.2C. Still worked just fine.

There is one good way that is within 5% using lithium batteries, and that is Coulomb Counting (amp hours in and out) which is what all EV manufactures use along with resting voltages.

I had a Leaf for 7 years, and its minimum error was about 30%. One of its favorite "tricks" was going from 20 miles of range to 0 in about a quarter mile. It did coulomb counting via a Hall sensor.

Looks like you put your foot in your mouth again.

[Sunking]

You seem to have a lot of experience with small EV and golf cart packs, but not much with residential-scale ESS. The problems overlap, but not entirely.

At last count over 4,000,000 watts and counting in commercial and residential. How much have you got? FWIW there is no difference between an EV battery systems and Solar. Both charge and discharge at the same rates. EV just use much better BMS and batteries the residential consumer does not have access too.

[nebster]

Thank you just agreed with me and backed me up. 10% is horrible accuracy. Anything over 5% is unusable.

The new estimators appear to be able to achieve significantly better than 10% using a learning algorithm. They fuse coulomb counting, voltage, and the other inputs I mentioned earlier.

I do not have one yet, so I cannot personally attest to their performance. But I believe it is possible.

Here is the problem most folks do not understand and how you are nebster are spinning numbers. That 2S pack for the phone discharges at very low rates on the order of C72 to C/100 which for all practical purposes is Open Circuit Voltage.

No, this is incorrect. Under even a small discharge, the voltage of a lithium chemistry cell will be much lower than under asymptotic OCV. Likewise, under even a small charge, the voltage will be quite a bit higher.

What matters for voltage-based estimation is that the load (current) is stable enough for long enough that the voltage can begin to approach its stable, asymptotic point at that current.

The new estimators account for this, using the current-based data to refine the coulomb-counted data when the current is reliable enough, but disregarding it when it is not.

There is one good way that is within 5% using lithium batteries, and that is Coulomb Counting (amp hours in and out) which is what all EV manufactures use along with resting voltages.

Because counting always exhibits some drift, and because that error accumulates over time, that estimation technique can readily be off by more than 5%. It is only good for a short while, and then it is not so good any more. That is why it must be "resynchronized" regularly, and it is also why people are developing better estimators.

Particularly for those of us operating large packs at partial SOC for extended periods, the current-integrating approach becomes limited in its usefulness.

You seem to have a lot of experience with small EV and golf cart packs, but not much with residential-scale ESS. The problems overlap, but not entirely.

[Ampster]

You are full of chit KARRAK. SOC FYI is OCV. 10.5 volts is under HEAVY LOAD.

Karrak I got news for you. EV's DO NOT USE LFP CELLS.

No need to shout. Obviously someone has pushed your buttons. I trust his facts more than your cherry picked facts. Didn't you recently say that the only way to measure SOC was to disconnect from load for 24 hours. Heavy load creates voltage sag that skews the data. As far as I am concerned, despite your extensive contributions to this forum, you sometimes over reach in order to make yourself seem correct. I have lost confidence in what you often say and I do not have the time to do the research to ferret out the few morsels of valuable insight that you may still be capable of offering.
Note to self, ignore this poster.

[Sunking]

Note to readers: this poster cherry-picked his data here. He chose absolute 0% SOC voltage for LFP, but a much more reasonable 40% SOC bottom voltage for a lead acid battery. Lead acid at 0% SOC will measure roughly 10.5V.

You are full of chit KARRAK. SOC FYI is OCV. 10.5 volts is under HEAVY LOAD.

Karrak I got news for you. EV's DO NOT USE LFP CELLS.

[Sunking]

Yep. In 2004 I designed a 2S battery pack for a satellite phone that used state estimation. It took voltage, current, current history, temperature and age as inputs and provided SOC as an output. It was accurate to about 10%.

Thank you just agreed with me and backed me up. 10% is horrible accuracy. Anything over 5% is unusable.

Here is the problem most folks do not understand and how you are nebster are spinning numbers. That 2S pack for the phone discharges at very low rates on the order of C72 to C/100 which for all practical purposes is Open Circuit Voltage. So you you can get a SOC on a battery with extremely light loads, but 10% is horrible. You an Nebster also have not defined SOC. I understand why you do that because you are advocates and need to spin numbers so consumers do not understand.

There is one good way that is within 5% using lithium batteries, and that is Coulomb Counting (amp hours in and out) which is what all EV manufactures use along with resting voltages. You are only telling half the story. I can take any brand new battery, fully charged healthy battery and make it measure 0% SOC. Or I can take a dead battery and make it read 100% fully charged. Both are false readings at 100% error.

Why do you think so many come here and ask why their expensive battery monitors lie to them and give them false readings?

[jflorey2]

This is also no longer true. There are now state estimators that model recent historical current plus instantaneous voltage to calculate SOC. They take into account the recent load/charge dynamics that affect the voltage now.

Yep. In 2004 I designed a 2S battery pack for a satellite phone that used state estimation. It took voltage, current, current history, temperature and age as inputs and provided SOC as an output. It was accurate to about 10%.

[Ampster]

@Nebster,
Thank you. Both of your responses were very helpfully and articulated the information in a very useful manner.
Of course that is what one would hope to get out of a forum like this.
It is unfortunate that some posters feel the need to to be less than civil in their responses.
Happy Thanksgiving to all.

[nebster]

You cannot use voltage as SOC on a working system. To use voltage as SOC requires the battery to be open circuit and rested for 24 hours. Two conditions that you cannot do on a functional system. Who on earth is going to disconnect their batteries and wait 24 hours? See any problem with that?

This is also no longer true. There are now state estimators that model recent historical current plus instantaneous voltage to calculate SOC. They take into account the recent load/charge dynamics that affect the voltage now.

In a similar way, when my LFP pack is in reasonably steady-state discharge (stable at e.g. 500W, 2kW, 4kW, or 6kW), I can look at the voltage and by experience tell you the SOC within 5%. That a computer with a learning algorithm can do a bit better is no surprise.

[nebster]

Hate to keep beating you up, but that is complete nonsense.

It's not nonsense. He means that in real-life, practical terms, over the SOC range that is usable, the slope of the voltage curve is flatter. And it is -- markedly so. Let's do the real math.

The flattest Lithium Ion is LFP (LiFeP04), and in a 4S operation for 12 volts is 10 to 13.8 volts or a spread of 3.8 volts from 100% to 0% SOC.

The range of normal operation is going to be much closer to SOC among the range {10%, 90%}. We don't take cells down to 10V except maybe once during prebalance. Smart users don't take cells all the way up to the top of the charge envelope for the sake of longevity.

At 10% SOC, LFP OCV after discharge is roughly 3.20Vpc. At 90% SOC, 3.33Vpc. The slope of the voltage curve over the usable interval is (3.33 - 3.20) / (90 - 10) = ~1.6mV per one percent SOC. In a 4s 12.8V nominal string, quadruple that to 6.4mV per percent SOC.

A 12 volt Pb FLA battery is 12.6 to 11.9 volts or a spread of 0.7 volts.

Note to readers: this poster cherry-picked his data here. He chose absolute 0% SOC voltage for LFP, but a much more reasonable 40% SOC bottom voltage for a lead acid battery. Lead acid at 0% SOC will measure roughly 10.5V.

In the real world, most of us try not to use more than 50% of lead acid, and thus the usable SOC is {50%, 100%}. The corresponding voltages are {1.98, 2.12} per cell. The average slope over that range, then, is (2.12 - 1.98) / (100 - 50) = 2.8mV per percent SOC. In a 6s 12V nominal string, we get 16.8mV per percent SOC.

Thus, the LFP slope is 2.6 times flatter than the lead acid slope over the usable range.

(In the narrower middle range, between say 30 and 70% SOC, the difference is even more.)

[Ampster]

You cannot use voltage as SOC on a working system. To use voltage as SOC requires the battery to be open circuit and rested for 24 hours. Two conditions that you cannot do on a functional system. Who on earth is going to disconnect their batteries and wait 24 hours? See any problem with that?

No problem. I was referring to something like the meter on my GEM which worked pretty well when it had Lead Acid batteries. Clearly it would not give the precision by measuring open circuit and waiting 24 hours. It did give an indication however that it was time to charge before going for a run to the hardware store.

[Sunking]

Yes, you are correct the range is wider but I was referring to the flat part of the Lithium charge curve before the knee. If I recall correctly the Lead Acid discharge curve is more gradual toward the end and the Lithium is steeper. Isn't that why voltage is not as good an indicator of SOC in Lithium vs. Lead Acid? I am just comparing my miniscule experience with LFPs vs FLAs.

You cannot use voltage as SOC on a working system. To use voltage as SOC requires the battery to be open circuit and rested for 24 hours. Two conditions that you cannot do on a functional system. Who on earth is going to disconnect their batteries and wait 24 hours? See any problem with that?