Datacenter UPS Heads Home For Off-Grid Power Solution

The news sites seem never to be without stories of Elon Musk and his latest ventures, be they rapid transit tube tubes in partial vacuum, space flight, or even personal not-a-flamethrowers. Famous for electric vehicles, Musks’s Tesla also has a line of solar products and offers the Powerwall home battery power system. These are tantalizing to anyone with solar panels, but the price tag for one isn’t exactly a dream.

[Nathann]’s budget couldn’t stretch to a Powerwall, but he did have access to a hefty ex-datacentre uninterruptible power supply (UPS) and a large quantity of lead-acid cells. From this he built his own off-the-grid power in the cellar of the home. It’s not as elegant as a Powerwall, but it can power the house on moderate usage, so he claims, for up to ten days.

On one level the installation is more of a wiring job than one of high technology, but the logistics of dealing with nearly 100 lead-acid cells are quite taxing. The UPS takes four battery packs, each clocking in at 288 V. The cells are joined with copper straps, and the voltage and current involved is not for the faint-hearted. An accidental short vaporized a screw and a battery terminal; if this were our house we’d put fuses in the middle of the battery packs.

The batteries are stored on wooden pallets atop brick pillars in case the cellar floods. The basement installation now is ready for the addition of solar and wind-based off-grid sources. Maybe your battery power solution will be less hair-raising, but it’s unlikely to be cheaper. Meanwhile this isn’t the first such project we’ve seen, though others usually go for 18650 Li-Ion cells, the use of lead acid remains a viable and economical solution.

62 thoughts on “Datacenter UPS Heads Home For Off-Grid Power Solution


      >”industry experience indicates that a 4 to 7 year VRLA battery life is more likely, regardless of cell size or warranty claims.”

      General purpose VRLA batteries have a design lifespan of 6-9 years. The lifespan calculation in UPS use is based on 1-2 discharges per year and the cut-off capacity is defined as 80% of the nominal. After that, the cells are more likely to fail because of internal corrosion and broken connections.

      The batteries they throw out are already one foot in the grave.

    2. Quote from the paper that Hackaday’s censorship dropped out:

      >”A European study of over 1,000 installations, of various system voltages and cell capacities, containing about 35,000 cells concluded that VRLA batteries require replacement after 5 to 8 years of operation. The shorter lifetime (5 years) was associated with batteries operating at 110 V or higher system voltages. The longer lifetime (up to 8 years) was applicable to better quality batteries and those of lower system voltages. Absorbed Glass Matt-type cells demonstrated a higher failure rate than did gel-type cells. No single or systematic factor was determined to cause this short lifetime. Representatives of a major battery company recently provided an update regarding VRLA cells’ expected life. Four different VRLA cells were tested, including three different AGM types and one gel-type. It was concluded that 20-year class AGM cells actually have an expected life of 5 years at 25°C”

    3. I was under the impression that these batteries have a useful lifetime of 5 years, and then they have to be replaced.

      Five years MTBF, so an older battery may work for awhile, but it could fail at any time.

      Is that not true with these lead-acid batteries?

      1. With the more robust and well characterised battery chemistries, that would be Mean Time Before Frigging-around-with-it, be that giving it overvoltage pulses, topping off electrolyte, blasting it with reverse voltage, cycling it in specific ways or whatever..

        1. I don’t think that does much more than delay the inevitable by a few weeks. When the plates start to break down and corrode, they get uneven deposits of material, the support grids break down and distort in shape to where they may short out, and you can’t magically reconstitute them back with zaps of electricity. You have to rebuild the battery.

          1. When they’re probably rated to run a data center for 5 mins while the genny spins up, and they take 10 days to discharge in household use, I doubt they will be punished hard enough for major plate distortions, or fulfill your doom predictions of greatly accelerated wear per fractional capacity loss, because they are getting used way below max ratings.

          2. That’s what I was thinking. In my experience data centre ups are designed to a certain rated capacity and then deployed in that capacity ie the loads the batts are hit with is very high. In a home situation you can give them an easier life and reap a much longer life. We do this naturally as in when grid fails we minimise our use and that in its self should keep the cells cooler, the mechanical stresses lower etc.

          3. The plate distortion happens because the material gets removed from the plate and distributed back unevenly due to concentration variations in the cell, temperature differences, voltage losses within the cell, sulfate crystals forming while the cell was discharged… etc.

            It’s impossible to cycle the battery and have the metal move back atom to atom where it was before.

          4. > doom predictions of greatly accelerated wear per fractional capacity loss

            That’s a characteristic wear-out mechanism for pretty much all types of batteries. The depth of discharge merely changes how soon it will happen and what’s the decay rate. With a 10% DoD your original lifespan will be about 1000 cycles or three years of daily cycling, and the battery will be properly dead in another 2-3 years.

          5. Right, but if you jump off the deep discharge curve at 300 cycles, where it’s nominally 3/4 used up at that rate, and switch to a lower draw, on the 55% curve at 600 cycles, you’d still have ~200 of those cycles left in it, and at around 5 day per cycle, that would take a decently long time.

          6. You don’t discharge a lead battery for five days – the longer it stays in a partially charged state, the more it develops sulfation problems. You have to keep it topped up at all times and charge it up as soon as possible.

            A good deep discharge battery with thick plates can take thousands of cycles with just 5-10% DoD but these are not that kind of batteries. They’re designed for low cycling and long (ish) shelf-life.

            >”DataSafe® HX batteries must not be left in a discharged condition after supplying the load, but must be immediately returned to float recharge mode”

          7. I get it, all cells are doomed to maximum failure mode. *scary noises* none shall escape the ordained fate. Seriously do you go telling people their car is a write off because of a stone chip in the hood?

          8. Look, it’s not my fault that lead-acid batteries ARE terrible for any sort of long-term energy storage or use, even when they’re designed specifically for the purpose. You’re better off with just about anything else.

    4. Yeah the article just calls them lead acid which is entire category of types and not a single type.

      They look like plain SLAs all be it very large SLAs at 92Ah.

      They likely have been kept at float most of the time and not cycled much so there should be some life left in them yet as long as the depth of discharge is kept low.

      SLAs are often used in UPSs as the cycle count is expected to low so generally they die from age rather than cycle count or discharge depth.

      1. Depends. When you have lots of cells in series, some will experience excessive charge and others will be undercharged – the result is plate and connector corrosion, sulfation and oxidation, or the venting of water out of the cells, all of which will eventually kill them.

    5. There’s also the misconception that a battery that is “dead”, as defined by some cut-off point like 80% capacity, will continue to work at that reduced capacity with the same rate of capacity loss.

      This isn’t the case. If a battery lasts 500 cycles to 80% then it has lost 20% of the active material and the remaining material experiences greater current for the same (dis)charge energy throughput. The corrosion, the joule heating, out-gassing, etc. increase at the square of current, so the same load on the battery causes it to degrade exponentially faster.

      Following the decay, each 20% drop in capacity happens at approximately 500, 320, 200, 130, 80… cycles and the corresponding remaining capacity is 80%, 64%, 50%, 40%, 33%… In other words, it gets faster and faster to break down the battery until the capacity is just dropping like a stone.

      1. And for lead acid batteries, since they’re sensitive about the depth of discharge, having the same load will cause the DoD to become deeper every cycle, which also accelerates the breakdown.

        So when they give the used batteries to you, they’re really just having you pay the cost of carrying them to recycling.

          1. Yeah, about 27 cents a pound, which means a 22 lbs battery nets you about $7. The recycling center or scrap yard however won’t pay you half that because they have to make profit on it as well. With 96 batteries, you have about $300 worth of scrap, which is incidentally just as much as it costs to hire someone for a day to dismantle your UPS and take them away.

      2. – out-gassing would be one of my concerns for putting a battery bank like this in my house … Not so much the high-concentration buildup side, but just it not likely being great to be breathing in any concentrations on a long-term basis…

          1. I should probably research more first, but I thought ‘Sealed’ LA’s were really still vented for pressure buildup. I know more for sure I’ve seen SLA’s go rouge in a DC UPS producing some rather potent DC fumes (within replacement schedule), and maybe that was a rarity, but seems even more likely to happen in an off-label SLA use like this.

          2. SLAs are vented for over-pressure.

            They have internal catalysts which recycle the gasses back to water (hydrogen + oxygen = water) but the process is not instant and while they’re not converted back, the hydrogen diffuses out through the plastic and the oxygen corrodes the plates.

    6. Hi, thanks for your comments, It was very interesting.
      they are expected to be charged daily with a DoD betwenn 5% and 15% and from the graph you show the lifetime seems to grow linearly as the DoD drop. their currently at 80% of their initial capacity so I believe I should still be able to squeeze a few hundreds cycle ?
      Another datacenter must change them in two years so maybe I will get a few more years out of my little experiment.

      1. You basically have a cycle life and a calendar life.

        Keeping the cells on trickle charge will slowly oxidize the connectors and plates due to the generation of oxygen from the water inside the cell, but allowing the battery to remain undercharged will slowly deposit sulfur on the plates and disable them, and then dealing with the sulfation issue you have to overcharge the cells which again causes damage by oxidation and water loss. In other words, your batteries are slowly dying even when you’re not cycling them at all, and this is why the data centers must swap them out once in a while.

        The cycle life can be extended quite far by light cycling, as you are doing, but the cells will have become unreliable due to the accumulated damage (oxidation/sulfation/water loss) and they can go through various failure modes ranging from loss of active plate area, short circuits, open circuits, overpressure, internal arcing and hydrogen explosions… you mileage may vary.

        I would say, however many cycles they’ve done up to date, that’s how many more you can expect till they’re truly dead – with the caveat of random failure.

  1. In my experience it’s five years or less and it’s off to the junkyard, weather used or not. When they swell up it’s time. I wonder if a gravity weight generator would work? Oil the pulleys check the cable, kiss.

    1. In a home setting compressed air would be easier to implement. As all it needs is a space for the airtank(s) it can fit in any shape and size of space reasonably well. (Though I think one step further and liquifying air for better energy density would be a winner it comes down to how long and how much energy you really want to store)

      pumped water would also be a better option I would suggest – again you can fit the water tanks to pump up to in how every many spaces you like above the pump much easier fit in than a big lump of lead.

      1. Compressed air is only efficient if you can do it isothermally. Small air tanks are barely 10% efficient at storing energy because the gas gets really cold when it comes out and that loses you all the pressure. You need a very large tank so the compression and decompression happens slowly enough.

        1. While you are not wrong It depends hugely on pressure differentials, expected flow rate, among many other things just how efficient you can make a system.

          Though the efficiencies are never going to be superbly high a properly made CAES system could last almost infinite cycles so you gain in not needing to recycle the nasty chemical soup in batteries all the time. The overall lifetime efficiency of CAES is where it starts to come back into play. The other potential gain for CAES if built right is lack of self discharge (but that does add more engineering challenges)

          With a DIY in home setup being cheap, safe and enduring makes it a reasonable choice, depending on your available space (it will use up more space than any battery per W) and power requirements. I have contemplated having a short duration UPS for the whole house and roughly calculated that a compressed air tank in with the hot water tank (using the thermal mass of the water to help keep the temp and therfor pressures up) should provide more than enough for that role despite wanting to drain a small tank rather quickly.. But that was only looking to provide that 5-10 mins for an automatic safe shutdown of all the computers and running the odd LED bulb while the power was out..

          And overall retrofitting into any home a compress air system won’t be a huge structural worry as CAES systems are not as heavy, especially when compared to the pumped water hydro and solid masses for which you might have to rebuild your house to take the load (even more so as to work the mass has to be as elevated as possible where CAES can be sited anywhere and work.. So CAES makes more sense than echodelta’s suggestion for most. (Not that batteries are a bad solution either, reasons for and against them so do some research and then take your pick.)

          1. Efficiency is kinda the thing, because if you have a home solar system with a payback of 20 years, and you toss away 50% of the energy, then your payback becomes 40 years plus the cost of the battery. In other words, unless you have no other options, it doesn’t make any sense.

  2. My first thought is, very nice – but how long are all those second hand batteries going to last. This seems to be an issue that nobody talks about much with battery powered energy solutions. The batteries are a consumable.

    1. thats a question I will asses as time pass. We are planning our energy source and we specifically choosed an Inverter for the solar panel that can use lithium battery. When the time as come we will install a battery upgrade on the solar panel Inverter and all we get in between is free beer.

  3. Unless that UPS is an online (ie. continously active inverter), will state the obvious here – given the audience – that UPS will fail, most datacenter UPS’es are only designed to support the load transfer to a generator.

    And hopefully his homeowners insurance is paid up. That’s one gigantic accident waiting to happen – seriously doubt the local AHJ would agree that any of it “meets code”.

    1. – Maybe a silly question and I’ve seen a limited set of UPS’s, but aren’t most all DC UPS’s online? – Most all I’ve seen are continuously online power conditioning – and at least the handful I’ve bothered to check, they can be cold-started without grid power even, so nice for something off-grid/solar type reuse if wanted (granted they don’t have provisions for charging without mains power built in – but not needing a grid sync to start is nice).

    2. This UPS was installed with no plan of firing a generator. It should have lasted around an hour in normal condition with full load. here we are talking about 5% of that load on average which I hope will keep the capacitor cool enough that they dont blow up too soon.

      What accident are you thinking of ? fire hazard ? blackout ? maybe I forgot a scenario … thanks for your reply

      1. Perhaps you should study this video. You have no safety mechanisms at all (at least none that are visible – DC disconnects, fusing, fire suppression, shielding of connectors, etc). Talk about “Darwin Award”….geez

  4. As a former UPS maintenance and repair technician, I’ll back up what everyone has said about a 5 year lifespan in service. In larger UPS system (which are the ones that would use the batteries depicted), they are indeed designed to run no more than 5-10 minutes while waiting on the generator to spin up. But also, if I had a string of batteries that suffered a short circuit like [Nathann] did, I would have removed every battery involved and replaced them. Once you short them out, they are no longer reliable enough to use to protect the load.

    1. My experience with old UPS batteries is that a single deep discharge can entirely destroy them. The UPS appears to “work” for as long as the power transients are very short, but the instant you have a proper blackout and you need the battery for more than a minute, it croaks.

  5. I got hold of a cheap second hand pc ups in my mid twenties. It’s battery lasted a lot longer when I swapped out the tiny moped battery for one from a lorry. Granted I had to make a few other alterations but it served me quite well for many years. I’d quite regularly get power cuts. It would happily run my PC with multiple monitors and my fridge freezer. Not quite enough for a hot plate though :D

    1. Exactly, the key here is to oversize the battery and minimise the load. If you did put that hotplate on it the life of the batt would have been much shorter due to the size of the discharge hit which would have caused some extra heating and thus warping internally. Oversized, batts and inverters stay much cooler, transients dont tax components as much if at all and thus live longer. The worst case is undersized installation.

      1. Technology matters too. Older SCR based UPS inverters have high losses and they heat up a lot, so they’re generally not even designed to run continuously. These things are built to cost, so replacing the small 20 Ah battery with a big 80 Ah truck battery just moves the failure point to the switches.

        1. Exactly. Hence having to do a few upgrades. The traces on the PCB had to be reinforced amongst other things. I scrapped it a couple of years back and used some of the parts for another project.
          If I recall I sanded off some of the solder mask and built up the trace thickness with solder and cooker earth wire. It was a bit of a Frankenstein’s monster with a CPU cooler strapped to the things that got hot. A few of the resistors were swapped out too. Metal clips were attached to pull away some more heat. Generally I only used it to get things saved on the PC and then run the freezer. It would keep going for aa fair few hours. Not pretty but functional. It used to live in aa large tool box :D

  6. That’s great to see people repurposing stuff like this. I have an old 3phase (240v) unit rated at 10kw per phase I am planing to put in my place.

    Been such a high capacity unit it will cope happily with the induction cooktop and dishwasher etc with ease which things like the power wall and other battery solutions fall way short.

    I still need to sit down and work out how to allow the batteries to only charge when there is sufficient solar generation

  7. If your charge controller is half decent, it will include an “equalisation” cycle every 30 days or so – allow overcharging for a hour to bring the cells back in line. It will also use PWM charging to reduce the growth of sulfate crystals.

    I’ve got 1320ah of 2VDC deep-cycle lead-acid cells downstairs. They were installed in 2010, and they still provide adequate energy to run the house after the sun goes down. That’s over 3650 cycles – the charge controller won’t let them go below 20% DoD without cutting out and sounding an alarm. I *could* get it to auto-start the generator, but it’s never happened.

        1. Well, substitute with three 22 Watt LED bulbs or whatever, but it’s still a very small amount for a house.

          For example, a normal fridge will consume about 500 Wh in 8 hours, so you don’t even have the power for the light bulb.

  8. Its a valiant effort, and you will learn a lot… but the batteries will soon die, some sooner than later. Great blocks too… Enersys DataSafe 12HX380… But all you need is one cell in a jar to go “open circuit” to remove a 288V string from contributing. You have 4 strings in your UPS, so if you can measure the blocks with a battery tester like Midtronics, Megger, Alber, you could group the best to make 2 or 3 “good” strings. However, the statistics will make this painful as failures will continue. Hence why lead battery customers get to a point where it is no longer viable to replace a block here and there and they replace it all.

    Important to consider for home use is the cost of operating that UPS. Nothing is free, including the energy just to turn it on and keep the batteries charged, fans spinning and logic operating. Most are double conversion, fully on-line, so there is some efficiency penalty. At home loads, the operating power can add up… say it is 200W (?) or your unit… that’s 4.8kW per day for utility outage insurance. If the UPS has an off-line mode, where it only turns on power conversion during an outage (Eco mode… YMMV), that can save you some power.

    Measure a sampling of the blocks, and set aside the best for second life use in campers, lawn tractors, and smaller 12 or 24V based UPS or best are “inverter-charger” units that RV and off-grid folks use. I’ve repurposed 5 year old UPS blocks of the U1 size to work another 10 years in lawn tractor service. I supplied my neighbors too for which they are still thanking me. That kind of life just doesn’t happen with wet cells commonly sold in *Mart stores.

    Do check for good info.

    Good Luck and happy learning!

    1. Lead acid cells have a certain self-discharge as well, so you’re not home free with an offline mode either. If you’ve got a 100 kWh system and lose 5% a month, you still have to charge it up every month and run a bit more to equalize the cells, so easily 10 kWh per month lost there.

    2. thanks ^^
      I know I may have to replace some batterys over time. from what I learned in the UPS manual it should still work with only two strip of battery running and my time spent checking them and swapping them is “free”
      It do have an Eco mode with 98% eff rated so should be that much (0.5 Kwh a day) but since it come from the futur solar panel installation it should not be a big deal.

  9. Wet Nickel-Iron [NiFe] or Wet Nickel Cadmium [NiCad] are the real way to go. Add distilled H2O when needed. Gell Pb cells are not meant to be deep cycled. Ask the railroad company

  10. Light-loading keeping them charged up are definitely the key to longer life for lead-acid batteries. On our little sailboat we have one Group 24 12v deep-cycle flooded-cell battery (the basic kind, with removeable caps on the cells) that has hung in for going on 12 years, because of constant solar topping up and light loading. We also have a deep cycle battery in our basement (that I took from the boatyard garbage because it tested OK) that runs a 12v sump pump if the power fails.

    Question – can you actually equalize sealed or gel lead-acid batteries? i thought that was only feasible with the basic flooded type, because it can cause boiling and outgassing.

    1. Gel batteries is a bit iffy, but VRLAs can, and should be run through equalizing charge once in a while. They have a catalyst inside that re-combines the hydrogen back to water.

      For the sealed types, you just don’t run so much current through.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.