19" rack battery options for DIY

winny

Member
Joined
Jan 28, 2019
Messages
78
Cell harvesting has been going on for years but I need to plan for next step, actually putting them into packs. Aiming for 50 kWh in a 14s 48 V system.

I see a lot of open packs on shelves. I'm thinking about fire-proofing and ease of expanding. Filling a 2 HE 19" box with 20x18 cells would look professional but adding more cells to the system (14s for 48 V) would take 20x18x14 = 5040 more cells for each step. Cost of boxes and connectors would certainly add up too.

I made two drawings, packing inside a rack box and series-parallell connection if boxes to make 14s3p (packs, not cells). I didn't find an obvious way to make a Y-split with SurLok connectors, so I just put two on each box with internal busbar. It would take more cabling to make a junction box for each in a separate 1HE box but would be far more flexible.

Does my idea that packing cells into a steel box, preferably without ventilation and sealed, would limit the spread of a fire in case of an internal short in a cell somewhere hold any merit?

Does anyone have any similar system or have looked into it? Would you have done it differently in hindsight?

Would it be better to build each box for 48 V with internal BMS and balancing, same as Pylon? Current would certainly be lower but balancing would also take more effort (more BMSes).

19inch_box.png


rack_assembly.png
 
Does my idea that packing cells into a steel box, preferably without ventilation and sealed, would limit the spread of a fire in case of an internal short in a cell somewhere hold any merit?
tbh, not really. Lithium fire would be hot enough to melt anything you can affordably put them in. The only thing that might withstand would be a firesafe.
The only sure way to keep anything from burning down in case of meltdown is putting the structure that houses the battery not near any structure you care about, ie house, garage, workshop, etc.

Also, sealing the case without proper ventilation would cause temperatures to rise, unless you had some kind of heatsink/radiator built into the unit to suck the heat away.

Would it be better to build each box for 48 V with internal BMS and balancing, same as Pylon? Current would certainly be lower but balancing would also take more effort (more BMSes).
Honestly here, I think each shelf would be an independent system pack. So each one would be 48V capable. And yeah, this would mean more BMS's to hook up. But the management and monitoring would be soooo much easier to deal with. And if you did need to replace or working on a pack, it would be easy to detach it from the system w/o loosing the whole system due to a shutdown

With that, you would want to make each battery with either front lugs, like you have in the drawing, or make each one with socket connectors that when you push the pack into its home it makes all the connections. Kind of like the way a blade server rack works. Tho this takes a little more planning out to make sure alignment is dead on.
If going with the above design with the terminals on the front, then you'd want to also make a common bus bar connection point on the front that all batteries/packs connect to in parallel. This will ensure that each pack gets an equal charge/load and minimizes overheating.

David Poz did a test on this and found that common bus is better than daisy-chaining the packs.


With that much space per pack, I would also add a little extra room between rows to allow for any heat dissipation. You would also want to add several temperature sensors throughout the pack to monitor for any hot spots.

I would imagine the graph to something like that would look something like this:
1669206385357.png

This is actually the temperature points in my server. Lots of sensors :lol: But you wouldn't need anything quite that elaborate.

I actually had thought up something similar to what you are planning. Hence the thought about a connector in the back like a blade server. I just never got a chance to get working on it or expand on the idea. Just residing in my head atm :p
 
Agree 100% with @Korishan. I have a large 18650 (lithium-ion) 108kwh battery bank with modest protection around the overall battery bank. But it's not physical protection that's top of my mind but rather these 3 things....
1) Cell level fusing
2) Low-stress operating conditions in terms of voltage range, ma/cell, and ambient temps.
3) Backed up with a BMS system to do shunt-trip disconnect if critical metrics are ever reached.
The number one, easily avoidable issue I see from many posts is that overcharging can definitely cause a fire to start - thus my emphasis on #2 and #3. #1 is the hedge against spontaneous individual cell failure that could allow dangerous (fire) level current within a pack.

I did go with Batrium BMS and it let's me see each pack day after day as a 'blue bar' on my computer screen. After 4 years of daily operation observing pack behavior I'm not worried about 'spontaneous' fire. I believe I'll see gradual behavior changes (as a warning) over many weeks or months before critical metrics are reached. And as mentioned above, the individual cell level fusing should mitigate in-pack fire.

**I'm only speaking for my own personal level of concern and of course you should make you're own decisions!! No judgements from me :)

Here's the Batrium status screen... easily visible if a pack get's out of whack.
1669219889500.png


Here's of the battery bank behind the blue bars above....
1669220094365.png




Here's the @DavidPoz youtubes on daisy-chain vs bus that @Korishan mentioned. This is Part 2.....
View: https://youtu.be/vUsHznMa76U
 
Last edited:
1) Cell level fusing
2) Low-stress operating conditions in terms of voltage range, ma/cell, and ambient temps.
3) Backed up with a BMS system to do shunt-trip disconnect if critical metrics are ever reached.
1. Yes, on both sides with the battery hookup nickel strip.
2. Aiming for 10 kW on a 50 kWh bank, so C/5. I expect very little heating from this.
3. Everyone seems to go for Batrium, but the website gives very little information and just about all links goes to the webshop. I don't want to buy and only then see how it functions. Going for 48 V per box would also mean I would need a lot of them, with about 30 separate 48 V systems? What other options do I have?

48 V per box would facilitate upgrading the system as more batteries are harvested.

Thanks for the video. I will build them myself so there won't be any cheating with copper plated brass busbars. I'm an EE so no issues with dimensioning either a daisy chain or junction system for it.
 
tbh, not really. Lithium fire would be hot enough to melt anything you can affordably put them in. The only thing that might withstand would be a firesafe.
The only sure way to keep anything from burning down in case of meltdown is putting the structure that houses the battery not near any structure you care about, ie house, garage, workshop, etc.

Also, sealing the case without proper ventilation would cause temperatures to rise, unless you had some kind of heatsink/radiator built into the unit to suck the heat away.
I was thinking about packing the boxes with some flame retardant like vermiculite or this one:
View: https://youtu.be/JAAxCxjJmDs


Aiming for C/5 so I don't expect much heat generation. I will have temperature sensor(s) embedded in each pack.
 
Last edited:
The biggest difference here is that LiPo packs react a lot differently than Lithium cylinder cells. 18650's and other cylindrical cells are like road flares or small rockets. So anything you put on them will need to have weight, or packed tightly on them, to keep the material there. You can't just cover them in the material and call it good.
So in your current design, you would definitely need a lid on top of each rack to make it enclosed. Then you need space for the vermiculite, probably about 2-3 inches all around clearance for a good barrier.

Not saying using it is a bad idea, just be aware that the metal cans will react quite a bit differently than the pouch cells ;)
 
3. Everyone seems to go for Batrium, but the website gives very little information and just about all links goes to the webshop. I don't want to buy and only then see how it functions. Going for 48 V per box would also mean I would need a lot of them, with about 30 separate 48 V systems? What other options do I have?

48 V per box would facilitate upgrading the system as more batteries are harvested.
I see 11 boxes in the pic above - 3 'batteries' with 3 boxes each and a 4th 'battery' with 2 boxes? It's not clear to me what a 'pack' (set of parallel cells) is in you're design. Maybe you can do 14 boxes in the rack - each box = 1 pack.

>Aiming for 50 kWh in a 14s 48 V system.
50kwh / 11 boxes = 4.5kwh per box. With 2600mah cells that would be around 14s45p per box or 630cells per box.

Can you share in more detail what you plan in terms of parallel / series cell arrangement in each box? How many cells in each box? Do you plan to expand over time?

Batrium is not the only solution for sure. I like the longmon orientation (in spite of huge $) because it's 1 central BMS system and you just plug in 14 additional longmons to add a battery. I'm at 112 longmons now but plane 42 more in next 2 years. Also longmons are useful because they can balance *large* (500ah+) packs reasonably.

But there are many choices . My point is that a good BMS is a key component to make sure a fire never starts and can be a source of metrics to monitor trends which can further mitigate risk. A larger battery bank - such as the 50kwh you plan - can benefit from a more nuanced BMS approach + you might find you want to expand...

Being a EE you might take to the @Stuart Pittaway (youtube channel) DIY approach which is similar to Batrium in terms of central monitoring with remote units. Here's an @adam Welch youtube showing an earlier version of Stuart Pittaway hardware to give you the idea. @adam Welch did several youtubes on building them.
View: https://youtu.be/DKz48WeSGlU

@Wolf (one of the senior members of this forum and a builder of a large 18650 battery bank) recently started a thread on building Stuart Pittaway's DIY BMS from scratch - https://secondlifestorage.com/index...ms-by-stuart-pittaway.11871/page-2#post-84938
 
Last edited:
For a good reference on how lithium battery fires work you should check out the firefighting guidelines that tesla teaches for fire departments. It can take as much as 40,000 gallons of water and sometimes a complete dunk of the car/battery in water for an extended duration to stop the fire. If battery fire is a concern for you, and it really should be given the stakes, then a separate, climate controlled building is the ultimate safety margin. Worst case you lose the battery shed and the house is safe. We build for safety and efficacy and while no one expects an issue it something that should be factored in accordingly.

I really like the clean rack builds. Something like that is my ultimate dream but it is a few years down the road for me at this point.
 
I see 11 boxes in the pic above - 3 'batteries' with 3 boxes each and a 4th 'battery' with 2 boxes? It's not clear to me what a 'pack' (set of parallel cells) is in you're design. Maybe you can do 14 boxes in the rack - each box = 1 pack.

>Aiming for 50 kWh in a 14s 48 V system.
50kwh / 11 boxes = 4.5kwh per box. With 2600mah cells that would be around 14s45p per box or 630cells per box.

Can you share in more detail what you plan in terms of parallel / series cell arrangement in each box? How many cells in each box? Do you plan to expand over time?

Batrium is not the only solution for sure. I like the longmon orientation (in spite of huge $) because it's 1 central BMS system and you just plug in 14 additional longmons to add a battery. I'm at 112 longmons now but plane 42 more in next 2 years. Also longmons are useful because they can balance *large* (500ah+) packs reasonably.

But there are many choices . My point is that a good BMS is a key component to make sure a fire never starts and can be a source of metrics to monitor trends which can further mitigate risk. A larger battery bank - such as the 50kwh you plan - can benefit from a more nuanced BMS approach + you might find you want to expand...

Being a EE you might take to the @Stuart Pittaway (youtube channel) DIY approach which is similar to Batrium in terms of central monitoring with remote units. Here's an @adam Welch youtube showing an earlier version of Stuart Pittaway hardware to give you the idea. @adam Welch did several youtubes on building them.
View: https://youtu.be/DKz48WeSGlU

@Wolf (one of the senior members of this forum and a builder of a large 18650 battery bank) recently started a thread on building Stuart Pittaway's DIY BMS from scratch - https://secondlifestorage.com/index...ms-by-stuart-pittaway.11871/page-2#post-84938
Sorry about that. My potato computer with slow clap processor gave me <1 FPS when I threw all boxes in the rack so stacking racks and daisy-chaining them wasn't an option CAD-wise. I'll simplify it and make a better overview. I got 1s360p per box in my attempt there with some room to spare on each side for fuse link, BMS and connectors intruding into the box from the front side. I need to seriously investigate going the other route for 48 V per box (14s24p) instead for ease of paralleling and expanding.

Thank you! I'll read up and contact them. Developing my own BMS is not a too large of a project for me. I design power supplies and do power system design at work on a daily basis, and I used to design battery chargers for a living. I have not yet seen anyone do a full-fledged BMS with automatic pre-charge, hybrid MOSFET+relay design to get best of both worlds and WiFi for telemetry to simply not have to deal with cables and isolation.
 
The biggest difference here is that LiPo packs react a lot differently than Lithium cylinder cells. 18650's and other cylindrical cells are like road flares or small rockets. So anything you put on them will need to have weight, or packed tightly on them, to keep the material there. You can't just cover them in the material and call it good.
So in your current design, you would definitely need a lid on top of each rack to make it enclosed. Then you need space for the vermiculite, probably about 2-3 inches all around clearance for a good barrier.

Not saying using it is a bad idea, just be aware that the metal cans will react quite a bit differently than the pouch cells ;)
Yes, separate building is plan A. But there isn't a hand grenade worth of energy in one cell failing so I doubt it will penetrate two 2 mm thick steel plates and go from one box to the next. Side note: for science/f-up at the job, I did once detonate a hand grenade worth of energy stored in capacitors at work. We have an explosion proof test facility with separate control room (about 15x15x6 meter) so no one was hurt but the door did buckle outwards a bit hand had to be persuaded back into shape to open and close properly. The 100k FPS camera went from sparks forming into pure whiteout from one frame to the next. I thankfully haven't seen any litium battery failure that energetic 🤯.
 
But there isn't a hand grenade worth of energy in one cell failing so I doubt it will penetrate two 2 mm thick steel plates and go from one box to the next.
It's not about a single cell going bad and flame-on. It's about it setting off all the other cells in the rack. Just do a youtube search for Tesla Fires and you'll see what a pack of cells going up in flames can do. There's not much left of the vehicle afterwards.
And it's not about starving O2 from the flames, either. Because the chemicals that the electrolyte is made from has Oxygen in it already. So it's self sustaining. The only thing that can be done to kill the flames is to remove as much heat from it as possible. Which, from a home owners view, is gonna be very difficult. Not like you can dump a full dewar on the pack(s).

The don't "explode" like a grenade. They explode like a rocket going off. So lots of energy being shoved in one direction for a prolonged period of time. And then the other cells get hot enough to ignite and you have a whole drawer full of rocket-torches.
 
My favorite (scary) illustration of a runaway fire is from @Rich Rebuilds. They significantly overcharged an exposed tesla battery which started the fire. Fire starts at 3:00-3:10 in the following vid. As it progresses you can see things shoot into the air (6:00+) as @Korishan is talking about. And of course it cascades creating a huge amount of heat.
View: https://youtu.be/WdDi1haA71Q

Best to avoid a fire in the first place :)
 
Last edited:
My favorite (scary) illustration of a runaway fire is from @Rich Rebuilds. They significantly overcharged an exposed tesla battery which started the fire. Fire starts at 3:00-3:10 in the following vid. As it progresses you can see things shoot into the air (6:00+) as @Korishan is talking about. And of course it cascades creating a huge amount of heat.
View: https://youtu.be/WdDi1haA71Q

Best to avoid a fire in the first place :)
Classic! Planning on a multi-tired approach to not let that happen in the first place.

Spent some time yesterday on simulating abusing a DC MCB with auliary shunt trip as both hardware trip (MCB itself), software trip (aux shunt) and current sense towards a standard Texas Instruments BMS. Possible, but the interfacing towards that IC might be too much for me as opposed to simpler BMS just for voltage readout and do balancing, current measurement, protection, WiFi and display in an Arduino. I don't like to run high current though PCBs if I don't have to. The DIN MCB does steal a lot of space behind the panel though...

Work in progress:
48v_box_empty.png
 
Back
Top