Lithium in UPS - updated with LFP

[BobWalance]

The small uninterruptable power supplies that I use have all needed their lead-acid batteries replaced periodically.

I decided to design and build the electronics necessary so that they could safely use lithium batteries.

Edit 3-31-2023: Version 4 of the circuit description is now for LFP only, plus the balance circuit has been simplified and is now much easier to build.

[BradC]

G'day Bob,

That's a neat pair of solutions. It'll be interesting to watch the LiPos age.

I have a pair of SmartUPS here, one with 72Ah and one with 127Ah @ 48V. I've often thought about how one would go about doing what you've done on a larger scale.

On the SmartUPS there is a heavy reliance on battery voltage to estimate remaining runtime and manage the shutdown procedures. That gets complicated when you have a system with a radically different discharge curve.

I'll be watching for updates.

[gnif]

You are aware that you can literally just drop LiFePO4 cells in place without having to change a thing? Only down side is you wont get full capacity from them due to the slightly lower charge voltage.

[ledtester]

The PDF says this about LiFePO4 cells:

Lithium Iron Phosphate (LiFePO4) was also investigated, but it has a maximum discharge rate of apparently only one amp of discharge current per ampere-hour of battery capacity (aka 1C), so a very large battery would be needed.

[BobWalance]

You are aware that you can literally just drop LiFePO4 cells in place without having to change a thing? Only down side is you wont get full capacity from them due to the slightly lower charge voltage.

The LiFePO4 chemistry would require different protection ICs on the BMS and Balance Board. I would have liked to have tried them, but their discharge rate is just way too low for this application. The huge battery size and associated expense wasn't tolerable.

[gnif]

You are aware that you can literally just drop LiFePO4 cells in place without having to change a thing? Only down side is you wont get full capacity from them due to the slightly lower charge voltage.

The LiFePO4 chemistry would require different protection ICs on the BMS and Balance Board. I would have liked to have tried them, but their discharge rate is just way too low for this application. The huge battery size and associated expense wasn't tolerable.

Fair enough, I have a set of 300Ah cells on a UPS here but just used off the shelf parts and not dealing with a huge amount of power.

[Psi]

You can get lithium batteries in common SLA form factors that are drop in replacements and also include the protection electronics built in.

They used to be really expensive and not worth it, but from doing some googling they look to be price more reasonably now.

[Geoff-AU]

The LiFePO4 drop-in replacements are the most obvious solution, but I have a cheap UPS here that started sometimes failing to switch to battery after replacing with a LiFePO4 pack.  Randomly it would work or not work.

The reason turned out to be that the UPS switches to battery at a random part of the mains cycle.  Sometimes this means the battery gets hit HARD as everything comes up to speed.  On a SLA it will just grizzle through at low voltage, current-limited by internal resistance, eventually catch up.  The LiFePO4 BMS sees the short circuit situation and shuts down.  My 7Ah pack was drooping down to 5V before cutting out.

I fixed it with a fist-sized capacitor, resistor for slow charging (when connecting the battery I don't want big sparks) and a diode for dumping its charge into the UPS quickly.  Now the voltage doesn't dip below 9V and everything is happy.

[Psi]

The LiFePO4 drop-in replacements are the most obvious solution, but I have a cheap UPS here that started sometimes failing to switch to battery after replacing with a LiFePO4 pack.  Randomly it would work or not work.

The reason turned out to be that the UPS switches to battery at a random part of the mains cycle.  Sometimes this means the battery gets hit HARD as everything comes up to speed.  On a SLA it will just grizzle through at low voltage, current-limited by internal resistance, eventually catch up.  The LiFePO4 BMS sees the short circuit situation and shuts down.  My 7Ah pack was drooping down to 5V before cutting out.

I fixed it with a fist-sized capacitor, resistor for slow charging (when connecting the battery I don't want big sparks) and a diode for dumping its charge into the UPS quickly.  Now the voltage doesn't dip below 9V and everything is happy.

Some small supercaps in series to get the right voltage would probably also work, and be less huge.
But if you had the cap already, makes sense to use it.

[TruslowPJ]

This is a pretty cool project.
I've actually swapped four UPS units with lithium batteries, but I went with LiFePO4 for a number of reasons. For me the most important factor was safety. UPS units get ignored and forgotten about until they fail, so I wanted a chemistry that can take abuse without becoming a hazard. LiFePO4 is much more stable and doesn't usually exhibit thermal runaway.
The second was voltage. 4s LiFe basically replaces lead acid as is. charging voltages are pretty much perfect. LiFe likes slightly more float voltage but the charge termination voltage is pretty much identical.
I have done a 12v and 24v unit the easy way, using NEC ALM12V7HPs which drop in as a replacement for 7 and 9ah SLA, with series strings up to 48v allowed and 60A continuous charge or discharge. my HP T1500 G2 has a 36v 12s4P A123 pack from an EMC standby power supply. I ripped out the BMS and put in a daly 60A. it doesn't fit inside but sits next to the unit.
my most recent project is a Tripp-Lite SU2200RTXL2UA. I used 16 Headway 38120 cells and a 100A daly BMS. This unit is a beast and I even fit all the cells in the battery slot with 3d printed holders, though I am still having trouble finding room for the BMS. at full load the current is only 34A, so maybe I could have gone with a 60A and had room for it alongside the cells.

I see your mention of wanting to use lifepo4 but discharge rates being an issue. Everything I used up until the most recent headway pack is using A123 cells, which were absolute beasts that could handle 60A continuously and 120A peaks. the NEC batteries were manufactured by A123 and the BMS could cope with quite a lot. it's sad that the drop-ins available now are mostly only rated for a 1C discharge due to using anemic BMS boards. I have a 4s3p pack of super old A123 cells back when they used a cardboard wrapper and it still is in good shape and I have seen it put out 200A starting my friend's miata.

the charging voltage of my SU2200 would also be perfect for 13s of 4.2v lithium cells, but I'm just not comfortable having >500wh of normal lithium batteries in my house.

your shunt regulator is especially interesting to me, as it could handle a charge circuit that was just a low float current and had far too high of an OCV.

[BobWalance]

...snip...

I see your mention of wanting to use lifepo4 but discharge rates being an issue. Everything I used up until the most recent headway pack is using A123 cells, which were absolute beasts that could handle 60A continuously and 120A peaks. the NEC batteries were manufactured by A123 and the BMS could cope with quite a lot. it's sad that the drop-ins available now are mostly only rated for a 1C discharge due to using anemic BMS boards. I have a 4s3p pack of super old A123 cells back when they used a cardboard wrapper and it still is in good shape and I have seen it put out 200A starting my friend's miata.

Lithiumwerks is producing some of the stuff that A123 had. They have 18650 and 26650 phosphate cells that look very promising. I will get some of these and see how they perform. The nice thing with LiFePO4 is that the 13.7V open-circuit from the UPS is just about perfect for a 4S configuration, so no regulator would be needed - just the proper BMS and balance circuit.

Thanks for the post.

[shapirus]

The LiFePO4 drop-in replacements are the most obvious solution, but I have a cheap UPS here that started sometimes failing to switch to battery after replacing with a LiFePO4 pack.  Randomly it would work or not work.

Interesting. In my UPS, in which I replaced SLA with LiFePO4, similar behaviour was caused by the BMS (of the Daly "smart" type) going into sleep after a period of time and not being able to wake up quickly enough on a sudden demand for current. The battery itself didn't (and still doesn't) have any trouble delivering it, and it's no surprise at ~7-8 mOhm per 32700 cell in a 8s3p arrangement.

It was solved by a cron job issuing a command (e.g., get SOC or cell voltages) over RS485 once in a while, which causes the BMS to never go into sleep.

[BobWalance]

I have built an LFP version of the UPS battery system and have updated the PDF (see the top of this thread).

The LithiumWerks 26650 LFP batteries (previously A123 Systems) are excellent in performance, safety and cost, and they will be what I use to replace the lead-acid batteries in all of my UPS.

Take a look at the updated PDF to see how they outperform the original 6Ah lead-acid battery. I really am impressed.

[BobWalance]

The circuit description document has been updated. The balance circuit is now much simpler. See the original post for the v4 pdf.