Battery Pack Sanity check

[BatteryManTakeMeByTh]

Hi all, I'm trying to source components to build my own battery pack. I need to get BMS recommendations too, but that's besides the point. My current design is 8p20s to get 72V and 160A out of 160 cells https://www.dnkpower.com/ur18650rx/ each at 20A rated output and 1950mAh capacity.

In short, what have I missed or am I not considering? Will this get too hot? Will I need thicker nickel strips than 15mm? I would greatly appreciate feedback so I don't die trying to spot weld everything together. I have an EE degree, so I know the basics, but not the specifics of batteries.

 

[TPEHAK]

8 cells parallel capable 20 A each is going to be 160A. If you are planning to use all 160A but less than maybe 20% of working time and the rest of the time it will be maybe 50% of that power then it will probably be OK. But if you are planing to use 160A 100% of the time then it might overheat because it is large battery with some cells located inside with no air to cool them and other heating cells around them.

That 20A 100% duty is for one cell with air around it and with no other sources of heat.

 

[BatteryManTakeMeByTh]

8 cells parallel capable 20 A each is going to be 160A. If you are planning to use all 160A but less than maybe 20% of working time and the rest of the time it will be maybe 50% of that power then it will probably be OK. But if you are planing to use 160A 100% of the time then it might overheat because it is large battery with some cells located inside with no air to cool them and other heating cells around them.

Thank you for that input. I definitely don't plan to be at 100% all the time. But I would like to implement some kind of cooling. If I'm doing math correctly, the capacity of the design is enough for 2 hours at peak (320Ah with 160 cells x 1950mAh per). Any ideas how I would cool something like this? Or is the general idea that you don't want to have to cool them?

 

[TPEHAK]

I do not know if there is an easy and cheap solution for cooling batteries. It might be cheaper and more effective just use high current tabless cells and build the battery with amperage margin without cooling or add some cells parallel (which might take even less space and money than cooling solution).

 

[Epithemeus]

Thank you for that input. I definitely don't plan to be at 100% all the time. But I would like to implement some kind of cooling. If I'm doing math correctly, the capacity of the design is enough for 2 hours at peak (320Ah with 160 cells x 1950mAh per). Any ideas how I would cool something like this? Or is the general idea that you don't want to have to cool them?

Math is off. At peak 160a you'll get around 5.7 minutes. 8 parallel groups at 1950mah is 15.6 ah. Only parallel cells increase the Ah, the series groups increase voltage.

 

[BatteryManTakeMeByTh]

I do not know if there is an easy and cheap solution for cooling batteries. It might be cheaper and more effective just use high current tabless cells and build the battery with amperage margin without cooling or add some cells parallel (which might take even less space and money than cooling solution).

So just more cells to make less current per cell thus lower temps? Makes sense to me, and doable. I have not built anything yet, so I have yet to test the power output or temps obviously. I am happy to add more parallels later to make sure temps stay low enough.

 

[BatteryManTakeMeByTh]

Math is off. At peak 160a you'll get around 5.7 minutes. 8 parallel groups at 1950mah is 15.6 ah. Only parallel cells increase the Ah, the series groups increase voltage.

This is what I needed. Thank you. I thought something about my math had to be off with the ridiculous numbers. Might buy more cells then, which would help temp and capacity. We shall see. Thanks again.

 

[E-HP]

This is what I needed. Thank you. I thought something about my math had to be off with the ridiculous numbers. Might buy more cells then, which would help temp and capacity. We shall see. Thanks again.

If you care about pack life, you won’t be fully charging or discharging the pack. So more like 4.5 minutes.

 

[docw009]

I would look for a better candidate cell. These are 2AH 18650 cells. 100A for two hours probably should be using big prismatic cells or something with big LiFePO4 cells..

Off the top of my head, if using Li-Ion cells, I think you're looking at 20S-50P in 21700. Those would be 5AH cells, and you could run 2 amps per cell for 2 hours.. Maybe go up to 100P if you're doing this often,

 

[E-HP]

I would look for a better candidate cell. These are 2AH 18650 cells.

Agreed. A 20s4p pack using P50B 21700 cells would provide 20Ah instead of 15.6Ah, 240A instead of 160A (continuous), and half the cells (half as many spot welds), while taking up 2/3 the area.

 

[amberwolf]

Given the current required, you might instead look into used EV modules from places like GreenTecAuto and the like. They're already built (well), and made of cells intended to supply these kinds of currents continuously, etc.

Probably be cheaper as y ou won't need to buy and learn to use all the tools to test, sort, match, and spotweld a bunch of tiny cells together, and have to fix mistakes / replace bad cells / etc..

You also won't have to buy more cells than you need just to get enough to be able to test them all and match them up to work properly together (since it's a fair bet that any 18650 cells you buy will have a wide range of characteristics among the ones you get, and will require testing and sorting to build all the parallel groups equivalent to each other before you series them up).

 

[BatteryManTakeMeByTh]

If you care about pack life, you won’t be fully charging or discharging the pack. So more like 4.5 minutes.

Yeah, makes sense. Don't want to wear it out too bad.

 

[BatteryManTakeMeByTh]

I would look for a better candidate cell. These are 2AH 18650 cells. 100A for two hours probably should be using big prismatic cells or something with big LiFePO4 cells..

Off the top of my head, if using Li-Ion cells, I think you're looking at 20S-50P in 21700. Those would be 5AH cells, and you could run 2 amps per cell for 2 hours.. Maybe go up to 100P if you're doing this often,

This wasn't a spec goal thankfully, just a byproduct of my bad math and other specs. Probably shooting for 20-30 min full current at most. Don't want to have 100+ lbs of battery if I can help it.

 

[BatteryManTakeMeByTh]

Given the current required, you might instead look into used EV modules from places like GreenTecAuto and the like. They're already built (well), and made of cells intended to supply these kinds of currents continuously, etc.

Probably be cheaper as y ou won't need to buy and learn to use all the tools to test, sort, match, and spotweld a bunch of tiny cells together, and have to fix mistakes / replace bad cells / etc..

You also won't have to buy more cells than you need just to get enough to be able to test them all and match them up to work properly together (since it's a fair bet that any 18650 cells you buy will have a wide range of characteristics among the ones you get, and will require testing and sorting to build all the parallel groups equivalent to each other before you series them up).

I definitely agree it would be easier to do that, and likely cheaper in the long run, so I will look at them. However, part of the fun of the project here is the difficulty of doing what I can myself. I'll definitely have to go through and extensively test everything, but that's part of my job day to day, so that's fine. I appreciate the advice though. I will definitely check out GreenTecAuto. Any other recommendations of places to look?

 

[TPEHAK]

https://www.youtube.com/results?search_query=welding+battery+cells

 

[amberwolf]

Batteryhookup sometimes has some, but their deals are not as good because GTA sometimes has free shipping on their deep-sale items. And GTA specializes in automotive EV packs, including those intended to be used to fix those EVs, where BH has all sorts of other stuff and doesn't necessarily test what you're getting, etc, and you have more checking to do to see if their stuff is suitable for your project).

It can be fun to do it all, but there are times when using a premade bit makes something work so much better / safer / consistently that it is worth doing. A few points about that:

One of the problems with buying individual cells from *anywhere* is that it's virtually impossible to get ones that are better tan "close" to each other in properties, as far as matching them goes. Even poeple here that have gotten boxes of cells that appear to be "right from the manufacturer", all the same manufacturing batch, etc., have reported significant variation in properties. Whether that is in the cells or from their testing methods, we can't really know, but if it's the cells it makes it tough to build a well-matched-cell pack from, and that makes it harder to have a pack that starts and remains consistent (balanced) in it's cell properties throughout it's lifespan.

The closer the cells start out, the more likely it is they'll stay that way. (the EIG cells I'm using are perhaps a decade and a half old, and remain balanced to within a hundredth of a volt (best my Fluke 77-III can tell me) even now that their capacity and current delivery capability has degraded significantly--they all still are *the same* so no balancing or other bodgery is needed. (I don't even use a BMS, though I probalby should at least monitor at cell level, but I don't charge full (4v/cell) and I don't discharge deeply (usualy recharge after every ride, rarely lower than 3.7v/cell)).


Also, a pack built from many small cells has many more points of potential failure and variation, because there are that many more interconnects between cells, and within the cells themselves from inside the can to outside, and every one of those will vary in resistance (even if just slightly), and every one of them has a chance (even if tiny) of failing mechanically (or even being left disconnected in the first place, and undetected in the end product). The fewer parts, the less things there are to fail, so while the chances of any particular one failing appear to be greater, the likelihood of being able to easily tell something is wrong before further problems develop from the increased stress on the remaining parts is that much greater as well.


(I had another point but JellyBeanThePerfectlyNormalSchmoo said she had a balancing problem with her tummy that needed to be fixed, and coming back to the computer I don't recall what I was going to say).

 

[amberwolf]

This wasn't a spec goal thankfully, just a byproduct of my bad math and other specs. Probably shooting for 20-30 min full current at most. Don't want to have 100+ lbs of battery if I can help it.

30 minutes at 160A is 80Ah using the entire capacity of the cells involved. Assuming you don't want to use the full capacity of a cell to keep stresses down, and say you use only 80% of it (leaving 10% at either end) that would be about 100Ah of cells. If you want to leave 20% at either end it's more like 120Ah needed, I think.

The 52v 40Ah (14s2p) of EIG cells I have here in SB Cruiser are about 35lbs (more, I think) without any casing. Cells have improved in density since these were made, but it's still a heavy pack to get that capacity.

If you need 72v, that adds more than a quarter again of mass right there, so let's call it 45lbs (rounded up for worst-case estimation), assuming the same cells I have to get 72v 40Ah. To get 120Ah, you'd triple that weight, for 135lbs just for cells and interconnects, no casing, etc.

Let's assume cells ahve improved by a third, making a new pack only 2/3 of the weight that my old stuff was capable of (when new), that's still about 100lbs for just the capacity to do what you're after (assuming a lot of things, so variance from actual could be high).


Another effect to account for is that for a given power, you need current and voltage. If, especially at a low SoC, the cells sag greatly in voltage for the current being drawn, you may either not get the watts you want out of it required to do the job, or the voltage may drop so much that the system speed capability drops below requirements at that point***.

***whcih is a good reason to overspec the voltage required so the system isn't dependent on pack SoC for max speed, but is instead limited in speed by the control system.

 

[amberwolf]

It can be fun to do it all, but there are times when using a premade bit makes something work so much better / safer / consistently that it is worth doing. A few points about that:
<snip>
(I had another point but JellyBeanThePerfectlyNormalSchmoo said she had a balancing problem with her tummy that needed to be fixed, and coming back to the computer I don't recall what I was going to say).

More thoughts I remembered:

It's easier to make a few high-current interconnects than a bunch of smaller ones, especially if you don't want them to heat up.

If you're using larger cells already built into a module, you also don't have to worry about using interconnects that have to be thin enough and high enough resistance to be able to spotweld them, so there is less likelihood of them having to deal with heat in the interconnects.

(any heat in the interconnects is itself wasted power, meaning wasted capacity, that you would need to account for by adding extra capacity (and weight, etc) to the pack to compensate for. It also means there is voltage drop in them, further decreasing the available voltage to the system in addition to the voltage drop inside the cells, at currents high enough to create this drop).

Cylindrical cells will probably have a greater pack volume than pouch cells, even considering the compression hardware for the pouches, for a given capacity / etc, because of the airgaps between the cylinders, but it depends on the specific pouches used and the actual construction of either type of pack. The fewer cells used in either case, the less wasted space, but it should be easier to avoid wasted volume with the pouches, depending on the shape of the space the pack has to fit into.


Also, picking cells so that the current you will draw, worst-case, is well below what they are claimed to be capable of, would stress them less, making them last longer, and also cause less voltage sag under load for a given cell.

 

[amberwolf]

72V and 160A out of 160 cells https://www.dnkpower.com/ur18650rx/ each at 20A rated output and 1950mAh capacity.

Let's say you use cylindrical 2Ah 18650 cells. Each cell probably weighs around 45g. We'll just toss in 5g of interconnect for every cell (it's probably signifcantly more). (if you find significantly lighter cells they are probably either not as-claimed for properties (or are entirely fake) or the spec given is wrong).

We guesstimated previously you need perhaps 120Ah to get 160A for 30 minutes. So you need 60p of 2Ah cells for that (more really, because those aren't quite 2Ah). At 72v, 20s, that's 20 x 60 cells, which is a total of 1200 cells.

1200 cells x 50g = 60000g, or 84kg, which is about 133lbs. That doesn't include any of the casing, cell separators/holders, etc., and probably doesn't fully include the interconnect weights.

I haven't tried to calcualte the volume this pakc would take, but I would bet that it is significantly larger than the same done with the EIG cells I have, which IIRC are something like 7"x11"x0.25" in their holders. It has the advantage that you aren't limited to cubic types of pack shapes, of course.



Side note: the Sanyo UR18650RX hasn't been tested over at lygte-info.dk, but a number of other Sanyos have, if you're interested in comparing. The ones I glanced at don't have tests higher than 15A, but that's better than what the DNKpower site has for the "20A" cell linked; they only tested it up to 10A; based on how much greater the 10A is vs 5A, the 20A is probably *much* worse sag; at least twice as much. If so, I'd guess they'll sag down to 3.1 or 3.2v (or worse) at 50% SoC.

If that's true, then you'd only have 20s x 3.1v = 62v available at the 160A draw, when the pack is half full (about 3.65v/cell unloaded, or 73v). Assuming the same level of sag at full charge, then instead of about 84v (unloaded) you'd have about 0.55v (cell sag) x 20s = 11v (pack sag); 84v - 11v = 73v, so it would be like having your pack only half full wehn it's really full, at that current draw with that amount of sag.

Without actual test data I couldn't tell you what you'd really get, of course, but it's safer to assume worst-cases and then anything that turns out better is a bonus.


copied from the DNKpower page:

 

[eMark]

(I had another point but JellyBeanThePerfectlyNormalSchmoo said she had a balancing problem with her tummy that needed to be fixed and coming back to the computer I don't recall what I was going to say).

Probably a protective balancing BMS suggestion?

So what balancing BMS (if any) is on your 52v 40Ah 14s2p EIG (tummy). Assume you may be your own BMS?

Bottom line: What is your controllers' amp cutoff setting? How many minutes of continuous max discharge at near that rate would you consider reasonable for safety and of course cycle life longevity ?

 

[amberwolf]

So what balancing BMS (if any) is on your 52v 40Ah 14s2p EIG (tummy). Assume you may be your own BMS?
Bottom line: What is your controllers' amp cutoff setting? How many minutes of continuous max discharge at near that rate would you consider reasonable for safety and of course cycle life longevity ?

I don't understand the question...what is "tummy" for? (if you're referring to my comment about JellyBeanThePerfectlyNormalSchmoo's tummy, she doesn't have a battery in there (I hope; it would probably be problematic for her) couple of example videos of the schmoo.

Doesn't really matter, since this thread isn't about my stuff, and I don't need to change anything in my system; my replies are all to the OP. You probably want to talk to the thread OP instead.

 

[BatteryManTakeMeByTh]

Let's say you use cylindrical 2Ah 18650 cells. Each cell probably weighs around 45g. We'll just toss in 5g of interconnect for every cell (it's probably signifcantly more). (if you find significantly lighter cells they are probably either not as-claimed for properties (or are entirely fake) or the spec given is wrong).

We guesstimated previously you need perhaps 120Ah to get 160A for 30 minutes. So you need 60p of 2Ah cells for that (more really, because those aren't quite 2Ah). At 72v, 20s, that's 20 x 60 cells, which is a total of 1200 cells.

1200 cells x 50g = 60000g, or 84kg, which is about 133lbs. That doesn't include any of the casing, cell separators/holders, etc., and probably doesn't fully include the interconnect weights.

I haven't tried to calcualte the volume this pakc would take, but I would bet that it is significantly larger than the same done with the EIG cells I have, which IIRC are something like 7"x11"x0.25" in their holders. It has the advantage that you aren't limited to cubic types of pack shapes, of course.



Side note: the Sanyo UR18650RX hasn't been tested over at lygte-info.dk, but a number of other Sanyos have, if you're interested in comparing. The ones I glanced at don't have tests higher than 15A, but that's better than what the DNKpower site has for the "20A" cell linked; they only tested it up to 10A; based on how much greater the 10A is vs 5A, the 20A is probably *much* worse sag; at least twice as much. If so, I'd guess they'll sag down to 3.1 or 3.2v (or worse) at 50% SoC.

If that's true, then you'd only have 20s x 3.1v = 62v available at the 160A draw, when the pack is half full (about 3.65v/cell unloaded, or 73v). Assuming the same level of sag at full charge, then instead of about 84v (unloaded) you'd have about 0.55v (cell sag) x 20s = 11v (pack sag); 84v - 11v = 73v, so it would be like having your pack only half full wehn it's really full, at that current draw with that amount of sag.

Without actual test data I couldn't tell you what you'd really get, of course, but it's safer to assume worst-cases and then anything that turns out better is a bonus.


copied from the DNKpower page:
View attachment 377122

Definitely a good work up of the math. I realized I would need an inordinate amount of cells for this if I wanted that kind of capacity. I'm ok with something small for now, as this is more for the project than anything. I definitely didn't think enough about the sag though, so thanks for that as well. I'll take a look at some of the testing results. Seems like I'll probably just want higher capacity and lower current cells. If I go through with making a pack with the sanyo cells I have, then I'd never end up using peak current from any of the cells at least. We shall see what I can do.

 

[E-HP]

then instead of about 84v (unloaded) you'd have about 0.55v (cell sag) x 20s = 11v (pack sag); 84v - 11v = 73v, so it would be like having your pack only half full wehn it's really full, at that current draw with that amount of sag.

So with that voltage sag at 160A, power going to the motor is reduced by 14% (73x160 vs 84x160), or 2kW less power. You can feel a 2000W difference pretty easily.

Definitely a good work up of the math. I realized I would need an inordinate amount of cells for this if I wanted that kind of capacity. I'm ok with something small for now, as this is more for the project than anything. I definitely didn't think enough about the sag though, so thanks for that as well. I'll take a look at some of the testing results. Seems like I'll probably just want higher capacity and lower current cells. If I go through with making a pack with the sanyo cells I have, then I'd never end up using peak current from any of the cells at least. We shall see what I can do.

If you built the pack with P50B cells (capable of 50A discharge, like in the example I mentioned above) then a 20s4p would have 20Ah of capacity, and a pack capable of providing 200A discharge instead of 160A. Generally, more headroom equals less sag (and less heat as well). Selecting the right cells to match your requirements is your most important decision. Plus using the 21700 cells you'd require half the welds of a 20S8P pack of the Sanyos and you'd have 20Ah vs 16Ah of energy, and take up less space.