Multiple UPS Shared Parallel DC Battery Bank

Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.

Layla

Game Engine Developer
Jun 21, 2016
215
177
43
40
Hi Folks,

I'm thinking about trying to load-balance/share the battery banks of my multiple separate APC UPS.

The idea here is that the load against each UPS is not the same during an outage, and I'd like to maximize uptime for all my devices by pooling all of the batteries together, rather than having a separate battery bank per UPS.

In the DC power domain, this is relatively simple, just parallel each 120V DC battery bank, so you stay at 120V DC, but share the load across all batteries.

For example, if you have four UPSes, and each UPS has 10 x 12V batteries in series, you will now have say 4 sets of 10 x 12V in series, with each set in parallel, so 40 12V batteries in total, still maintaining 120V.

In this case, the DC chargers are also attached in parallel, which, since all the batteries and chargers are identical, shouldn't be an issue, either (though it could be if the chargers try to be too intelligent).

BUT, the catch is that there are a couple of APC proprietary "temperature/data" wires for each battery pack. I assume that if these aren't connected to satisfaction, the UPSes may do things like turn on the fans at inappropriate times. One idea I have is to run each set of temp sensors to a separate UPS (while paralleling everything else).

But, the UPSes do already support up-to 10 external battery banks each, so I sort of wonder if the "data" lines can already just be combined naively (or whether there's some extra complexity where the UPSes transmit the max(N) battery temperature, or something like that.

Really, I'm just wondering if anyone else has already done this before? It seems like a fairly obvious approach to improve load balancing and reliability if you have multiple separate UPS units, but everyone seems to opt for much more complex approaches where they either run UPSes in series or more rarely parallel on AC power, including synchronizing the sine wave outputs with one UPS being master and the others slaving to that. Much more complex stuff, and I'm not sure why people prefer to combine power in the AC domain instead of the DC domain (there are some benefits, but they don't seem worth the complexity to me).

For anyone who will say "just connect 4 battery banks to your single UPS like APC wants you to do" -- sure, now you have a single 3000VA UPS with 4x the runtime. Doesn't solve my problem. I want 12000VA net (4 x 3000VA, each a separate inverter, not a combined 12000VA like if I had combined the inverter AC outputs like many have done) AND sharing that combined runtime because they all draw from the same batteries - so they'll all "die" at the same time, not die at different times depending upon their relative current draw.
 

cesmith9999

Well-Known Member
Mar 26, 2013
1,417
468
83
You would need a really good BMS system. as 1 bad battery could really tank your capacity.

You can do this, what is the benefit you are looking to get? if I have servers with 2 power supply inputs I would want (and my current system has) two UPS systems with separate battery banks. That way I have an option to take a UPS offline for maintenance. where in your case you would need to take the entire system down for maintenance.

You may want to look at this video from a off-grid solar guy
who is doing something similar to his house.

Chris
 

Layla

Game Engine Developer
Jun 21, 2016
215
177
43
40
You have N parallel inverters, so you can still take a UPS offline while keeping the others up, same as if they didn't share the big parallel battery system.

But you are quite correct that it increases the chance of a bad battery harming the operation of the larger system, but since each UPS already uses 10 batteries in series, I'm assuming they are fairly reliable. I guess time would tell whether that's a good assumption :)

For multiple PSU servers, the point is kind of moot as you say, since you can take advantage of that at the PSU level... But for single input devices, it starts to matter a bit more (though ATS is an option). Also, a nice quality is being able to combine DC power used by both the 120V UPSes and the 240V ones.
 

T_Minus

Build. Break. Fix. Repeat
Feb 15, 2015
7,625
2,043
113
Just go with larger batteries commonly used in off-grid, boats, etc... and then go with the charge controller\BMS they would use too.
Using UPS for this wouldn't be my design choice.
 

Mithril

Active Member
Sep 13, 2019
354
106
43
So, first off 120V DC is dangerous, besides being a high enough voltage to cause enough current to flow through you to kill you (like high voltage AC), DC arcs FAR easier. With household AC (or simulated) the voltage crosses zero multiple times a second, meaning any arcing has to re-establish each time, yes you can ionize the air and/or vaporize some metal into conductive plasma, but DC is far worse as once an arc is established, you have to increase the resistance (generally the distance is the easiest way), or disconnect the circuit elsewhere. So not only do you need proper fuses, you need arc detection/suppression.

That being said: Generally connecting Lead-Acid batteries in parallel is a one way trip to sadness, unless all batteries are prefectly at the same charge, temperature, all resistances remain equal etc you WILL get charge imbalance, and the batteries (packs) will try to charge the lower voltage batteries, this can result in a thermal runaway if you are unlucky.

The right way to do it is to either: have Mosfets/solid-state relays controlled by circuits that monitor the voltage and current flow (via shunt resistors or similar) and connect/disconnect as appropriate. Or, You have one larger capacity battery (bank) connected via sufficiently rated diodes/mosfets/SSRs.

Honestly considering the cost/upkeep of SLA you might look into going with Lithium Iron phosphate long term (single large bank feeding the 3 UPSes), and maybe SSRs wired to only connect when power fails to avoid cross charging, obv with the required protection.

Lithium Iron Phosphate has a similar nominal voltage and charge profile to SLA, looks like a lot of solar/DIY powerwall are moving to it, much better cycle count than "normal" portable Lithium Ion, better thermal tolerance, better behaved when things go wrong; the tradeoff is larger physical size (still more dense than SLA).

I'm actually in the planning phases to have a 48v Lithium Iron Phosphate bank that feeds into my 48V SLA bank my main UPS uses when there is a power outage. Related I'm planning to convert some of my more critical equipment to dual feed with direct DC-DC power from that bigger 48V bank, POE technically allows for 44-57V so I could, in theory, do a fairly simply "diode OR" for certain things, not a huge loss and far FAR less than DC-> AC -> DC
 
Last edited:

Layla

Game Engine Developer
Jun 21, 2016
215
177
43
40
So, first off 120V DC is dangerous, besides being a high enough voltage to cause enough current to flow through you to kill you (like high voltage AC), DC arcs FAR easier. With household AC (or simulated) the voltage crosses zero multiple times a second, meaning any arcing has to re-establish each time, yes you can ionize the air and/or vaporize some metal into conductive plasma, but DC is far worse as once an arc is established, you have to increase the resistance (generally the distance is the easiest way), or disconnect the circuit elsewhere. So not only do you need proper fuses, you need arc detection/suppression.

That being said: Generally connecting Lead-Acid batteries in parallel is a one way trip to sadness, unless all batteries are prefectly at the same charge, temperature, all resistances remain equal etc you WILL get charge imbalance, and the batteries (packs) will try to charge the lower voltage batteries, this can result in a thermal runaway if you are unlucky.

The right way to do it is to either: have Mosfets/solid-state relays controlled by circuits that monitor the voltage and current flow (via shunt resistors or similar) and connect/disconnect as appropriate. Or, You have one larger capacity battery (bank) connected via sufficiently rated diodes/mosfets/SSRs.

Honestly considering the cost/upkeep of SLA you might look into going with Lithium Iron phosphate long term (single large bank feeding the 3 UPSes), and maybe SSRs wired to only connect when power fails to avoid cross charging, obv with the required protection.

Lithium Iron Phosphate has a similar nominal voltage and charge profile to SLA, looks like a lot of solar/DIY powerwall are moving to it, much better cycle count than "normal" portable Lithium Ion, better thermal tolerance, better behaved when things go wrong; the tradeoff is larger physical size (still more dense than SLA)
I rebuilt 3 120v battery packs tonight (10x fresh 12v batteries in series per pack) with no arcs. Agreed that 120V in general is dangerous and DC has it's own risks, but, I think AC is much more dangerous given it causes your muscles to contract and if you're holding a conductor you can be unable to let go. Whereas DC tends to cause you to let go immediately. In any case, people shouldn't be working on electricity without precautions.

I'm not up-to-date on LiFePO4 chemistry, but I was under the impression SLA was still king from a safety perspective?

Regarding the SLA in parallel being a one way trip to sadness, how do APC reliably allow you to connect up to 10 external packs to a single UPS if this is the common outcome? I understand the theory, but I remain skeptical that it'd commonly cause large issues in practice given the slow float charge for days after any battery drain. Though clearly it is important to independently float charge each bank fully before connecting them in parallel, to avoid low impedance charging from one battery bank to another. Also important to keep conductor paths impedance balanced.

I'm inclined to measure voltage, charging current, and temperature of each bank (which the UPS does) and throw a Zabbix alert if any bank goes out of spec relative to the others - possibly then tripping a relay to disconnect that bank for safety inspection.
 

Brian Puccio

Member
Jul 26, 2014
70
33
18
41
I think AC is much more dangerous given it causes your muscles to contract and if you're holding a conductor you can be unable to let go. Whereas DC tends to cause you to let go immediately.
I thought it was the opposite?
 

Layla

Game Engine Developer
Jun 21, 2016
215
177
43
40
Any current in the right conditions can cause muscle contractions where you can't let go. But it's much more common with AC. The 'let go' threshold (the amount of current it takes to cause muscle contractions where you can't let go) is lower for AC (as is the atrial fibrillation threshold), and at the same time, due to its nature, DC is more likely to initially push you away from the shock. The 'sweet spot' where you aren't pushed away and are not able to let go is wider/larger/much more common with AC. In general 50/60Hz AC is very easily deadly to humans.
 

Layla

Game Engine Developer
Jun 21, 2016
215
177
43
40
AC also causes sweating which further reduces skin impedance. No current conducting across your body is safe, but 50/60Hz AC in the 100-250V range is much more dangerous than most people seem to expect (even at very low current levels).

Much higher voltages come with more risk of arcing and such, so high voltage (thousands of volts) DC gets incredibly dangerous, because high voltage DC arcs can persist for many seconds - but I doubt anyone would expect a high current high voltage source to be "safe" :)
 

Stephan

Well-Known Member
Apr 21, 2017
920
698
93
Germany
Let me briefly play the party-pooper. ;-)

Word from people wiser than me: Stay at 48V DC. A simple overload, triggering a wrongly chosen circuit breaker i.e. one that cannot extinguish HV arcs, might burn down the house. Phew.

I also find it more convenient to have independent "building blocks" like batteries, UPS, servers, firewalls, switches, etc. Key to reliability of a system is imho easy replacement should a single device fail. Also repairability and if you are so inclined, redundancy. This is STH after all. Connecting batteries together has serious drawbacks: Equalization currents, arcing, need for more fuses, you really need a BMS to find weak cells, balancing etc. pp.

If this were for an entire house to use LiFePO4 for solar energy storage, I would contemplate building a system out of components. Like three Victron Quattros for 3-phase 230AC power, feeding off of 3+ batteries, charged by Victron solar charge controllers. Still saves alot of money, especially if you can supervise (BMS) and maintain (balance, replace) cheap-ish cells from Aliexpress yourself. Batrium from Australia really is the Maserati of BMS. Wouldn't touch any of those Dali brand cheap BMS or similar, you will waste alot more time and money in the long tail with those. Or go modular with a Byd system.

But for server UPS, I would not stray off the building block mantra. Just too much hassle. Plus the fire hazard if you haven't done this a 100 times already and really know what you are doing.