Stranger, if you have 10 cells in parallel and remove 1, and you don’t demand more than 90% of the pack’s peak rated current (the car’s electronics can automatically limit peak power until the battery is complete again through the BMS sending a network message to the motor controller), it works. Source : I am engineer and I work on motor controllers. I haven’t done lithium battery packs, but this is a straightforward bit of napkin reasoning based on the limitations of DC power supplies.
EV packs are groupings in parallel wired in series.
I’m an engineer and I work on launch vehicles and spacecraft, and no, you can’t just remove a cell from a complicated modern battery power system and expect it to function reliably. Such systems are designed to monitor the performance of individual banks and balance load or shunt defective banks as necessary to keep the power system performing within requirements, but if you just start removing cells completely the system is going to go haywire and start shutting down into protected or diagnostic mode.
Well there is no reason for it to do that, and if you think about it, that’s a mission critical fault in your launch vehicle BMS. Do you want it to fail in a state that leads to no power supplied or to fail in a state where the failed cell is shunted out of the circuit? You can bluster, but I doubt at your level you’re actually on the team that works on the pack design or the firmware for it. (you’re probably at a forest level, and we’re talking about a detail for an individual tree) I know from a systems perspective the failure handling model you describe is not correct, and I know that electrically it is possible to isolate failed cells. I won’t claim to know more than you do about launch vehicle BMS, I’m just saying it seems likely my model is more correct.
There is a difference between shunting around a battery that is failing (providing low or no current, wrong voltage, or just impedance mismatch) and a completely missing cell (open circuit). In the case of a missing cell you actually want to shut the system completely down and go to the redundant system (which all critical aerospace systems will have) because that is indicative of a catastrophic failure mode (potentially leading to fire or gross leakage of electrolyte) that the power system can’t even diagnose, and there is no way for a power system to differentiate between a removed cell and a broken circuit.
You are free to baselessly characterize my response as “bluster”, but while I haven’t written controller firmware (which is done by the contractor providing the controller) I have worked on design requirements for power systems as well as dynamic and functional testing of batteries, battery controllers, and power distribution systems in applications where fault tolerance and reliability are crucial to mission success, and failure doesn’t just mean pulling off to the side of the road and calling for warranty service but falling out of the sky or having to explain why a billion dollar satellite is now just another piece of useless orbital debris.
There was an episode of America Tonight where Virgil Simms (Jim Varney) showed off his electric car. Under the hood were hundreds of D cells and the show ends with Virgil taking the batteries out one at a time to find the dead cell.
Obviously it depends on the pack design, but a Tesla pack can handle individual cells failing either open or closed, and has high resistance to cascading thermal failures. The cells are arranged in bricks with ~70 cells in parallel each, so there’s little effect of a single failure.
Of course it’s impossible to ever remove a single cell. The packs aren’t designed for easy repair. Isolated failures are left in place.
Cells failing (generally means that they are not producing voltage within limits, but could also be high but finite impedance or no current) is not the same thing as a broken connection, short circuit, or missing cells. I’m morally certain that if the battery controller on a Tesla battery saw infinite resistance it would put the system into a failsafe mode (shutting down an entire section of the battery and probably putting the car in a service required “limp” mode), because there is no way for the controller to tell the difference between a broken connection/removed battery and a failure due to short circuit or battery rupture that could lead to thermal overload and fire.
One of the tests we do with new battery systems is intentionally underload or short them with feedback control disabled to see how long the battery can continue to function before it ruptures or explodes, which can be a delightfully energetic experience from a control room fifty feet away from the armored test cell. If this happened to a battery pack installed in a vehicle the result would likely be the entire battery pack catching fire and turning the vehicle into a molten pile of metal and glass slag. Lithium ion batteries like the 18650A cells in the Tesla battery pack can, once in thermal overload, develop prodigious flammability that is essentially impossible to extinguish, especially the NiCoAl cells they use for longevity and greater energy density.
Ok, so say the cell voltage is below the level of the other cells on the strip because of some failure. You disconnect it electrically - it cannot provide any useful work.
What, precisely, is the practical difference between doing that, and someone popping a door on top of the strip and removing 1 cell? I agree the BMS can detect a difference - infinite resistance like you said - but either way, the cell was not needed.
If there’s a missing cell but no heat measured from that region, oh well, right? Disconnect it electrically and the pack controller needs to reduce it’s current maximum by the portion supplied by the missing cell.
You’re telling me that if you have one of those aerospace battery packs on a bench, and you remove just 1 unnecessary cell - this might not physically be possible but the sensor could fail in a way that the pack controller thinks it did - it will just brick itself. Doesn’t matter if it was supplying current to the computers guiding the rocket during launch, and some other fault took out the backup.
Not a great design, and not a necessary tradeoff for a car. Tesla puts their cells in armored packs that can and do handle catastrophic fires. That is, the fire doesn’t reach the passenger compartment.
In any case, for the original topic, what I had in mind was if a cell fails (which is a near certainty if a pack has several thousand total), the car display just helpfully shows you where, and gives a barcode or QR code for the right part number. You either take it to a mechanic - and they scan it with their phone and order the part - or you DIY it. You disclaim all liability but design the pack such that it is at least feasible to reach failed cells without disassembling the car.
Maybe you would need to put a replacement in before you close the lid, and maybe the car could have a couple replacements already ready to go, kind of how there are extra fuses already in automotive fuseboxes.
How do you propose to remove a single cell from a sealed battery pack like the ones Tesla uses that are also liquid cooled or if you use conformal style packs like GM? We are talking thousands of cells. If you aren’t going to liquid cool how will you effectively keep the packs from thermal runaway and still reach the energy density regimes to have a useful range? Further, you’re going to have to make each bank removable in order to access dead cells in the centre (the most likely to die from thermal issues as they are the most insulated) and that means more structure which does nothing to keep the cost/ mass of the whole pack at a reasonable price level. It would make more sense to me to make a BMS that will inform the owner when the performance overall drops below a certain threshold and making the pack easily swappable than individually changing each cell which would be better accomplished at a depot level facility, just like the Renault Zoe.
FWIW, I am an Electronics Technician and former Avionics Tech and I get to repair ideas like you’re proposing all the time, where some brilliant solution is put in place but sideways and impossible to repair because it’s more elegant…:dubious:
It can’t do this on a cell level. The P100D pack has 8,256 individual cells, arranged into 16 modules, with each module having 6 bricks, and each brick with 86 cells in parallel. Total system voltage is 1663.8=~365 volts. The bricks are simple and don’t have any per-cell electronics or sensors.
I’m sure you’re right that it has sophisticated monitoring at both the brick and module levels and will disable the pack if anything serious happens. It certainly doesn’t rewire the pack to bypass entirely failed modules. But if one of the 86 cells in a brick fails, it depends on a thermal fuse, the cooling system, and “intumescent goo” to prevent further problems. One cell failing benignly wouldn’t be enough to take down the whole pack; the only way it could even be detected is as a transient, since there’s enough natural cell variation that you couldn’t detect an 85-cell vs. 86-cell brick in absolute terms.
There are EV battery styles that look like D cells. Pack 1, at the back of the frunk near the ground. Pack 2, under the passenger seat. Lift out the seat or frunk well. Have an armored fire panel on top. Unlatch it and remove it. Under it, each cell goes into a cylindrical well, with either air or liquid pipes around each one. Unclip the wire or tab from the top and push it to the side. Lift the cell out.
You could have the circuitry at the top of each cell include an LED so that when you do this (switch in the lid), the healthy good cells get a green LED and the bad ones are red.
There are a few issues. For one, if the terminals at the top of the cells are exposed and someone drops a wrench while doing this…
So you need to insulate the wires at the top. There are other mechanical issues with doing this and you need beefy enough current carriers and a more robust structure that allows you to change individual cells might in fact raise the manufacturing cost significantly.
This massively increases the cost and complexity. Take a look at a Tesla pack teardown sometime–the pack is essentially glued together, with bolts for good measure. This is good for environmental and fire resistance; it can withstand total water immersion for long periods, and contains internal fires long enough to pull over and safely exit the vehicle.
Removable panels would degrade this ability, or at the least weigh and cost more. Not to mention all the extra costs in being able to remove the cells (as compared to them just being soldered in place).
It’s better to just allow single point failures. Pack performance degrades slightly but not much else. If you have multiple failures in the same place, you’ve got bigger problems and there’s something defective/damaged with the pack itself.
The difference is that the controller shunting around a bank that has a misbehaving cell is hopefully avoiding any rupture or leakage problems, whereas if the controller detects the loss of continuity through a bank (which removing an individual cell will cause) it can’t know whether a connection is just broken, the cell is ruptured and venting, or there is a short circuit that is causing the bank to overload. You’d want to shut the system down, or at least the adjacent banks, to prevent cascading failures from the operating heat. Yes, the Tesla battery pack is designed to resist a presumed worst case thermal overload and rupture event without killing the passengers, but the minimum result of that kind of event is the unrecoverable destruction of an entire ~$30k battery, and potential significant damage to the vehicle. I take it by your comments that you’ve never seen catastrophic thermal overload of a lithium ion battery, but I can assure you that it isn’t just a little bit of smoke and flame. Here is a compilation of battery fires, and these are the relatively low energy density batteries used in cellular telephones.
Yes, if the controller on a battery operated power system for a rocket or satellite avionics detects a fault such as this, it will shut down the power system and switch over to the backup system. That is why all critical aerospace systems are at least double redundant, and in some cases use triple redundancy despite the weight penalty. That is also why we do all of this expensive dynamic environment and functional testing as well as insist on the “blizzard of paperwork” from suppliers regarding the provenance down to the component level to assure that components are appropriately acceptance or lot acceptance testing and are not counterfeit components from unscrupulous or careless suppliers.
I can imagine that Tesla, with their large, multibanked battery pack might just shut down the offending bank and shutdown or reduce power from the adjacent banks to put the car into a failsafe mode, but even that would result in a significant loss of range and power output. Removing a single cell will likely unbalance a bank, and there is only so much a controller can do to compensate for such a loss. In any case, the Tesla battery is a large sealed system with a highly integrated cooling system and various thermal sensors and I doubt that Tesla or any other battery manufacturer would welcome the liability of having someone untrained DIYer opening up the pack and start fiddling around or replacing cells with aftermarket cells of unknown provenance, much less design a complex battery pack to be field serviceable. This just isn’t a practical expectation from a liability, cost, or packaging standpoint.
The internal resistance of lithium batteries like the 18650 increases as they age; that is why they must be used and charged together – the batteries in a pack are ‘married’ – replacing one cell in a pack of multiple batteries is a no-no.
Although these batteries are now widely available and being used in consumer devices like flashlights and ecigs they were not designed for consumer use but for applications like locked metal boxes to provide battery backup for industrial machinery controllers where they would only be accessible to trained, knowledgeable people.
You still will have hundreds of cells even if you bump the size up from the ubiquitous 18650 (slightly larger than a AA) to something close to a D cell. Splitting the batteries of cells does nothing but increase the complexity and putting them where you suggest raises the C of G significantly. Most Li cells have some protection circuitry built in which prevents the issues I and others have raised. The battery on EVs are typically the single heaviest component of the car, and having the cells individually replaceable doesn’t make sense.
The idea of DIY is certainly admirable but to do so you sacrifice range, safety, and diminished cargo capacity while increasing mass, complexity and cost. That seems to me to be the definition of bad engineering practice.
There is another thing to consider in this discussion (which, BTW, has become a complete tangent from the question of the o.p. and probably deserves its own thread); rechargeable LiON and LiPo batteries in standard forms like the 18650A cell produced by reputable manufacturers and used in an approved fashion in a properly designed system are extremely reliable. The number of latent defects per million cells is probably in the single digits, so the potential for a single cell completely failing before its compatriot cells age out is pretty low. The fires and explosions you see on Youtube or sensationalized in the news are almost invariably because batteries are either used in a poorly designed open loop application where they can short out (vaporizers), overcharging or underloading, or purpose designed conformal packs used in devices intended to maximize slimness (smartphones and tablets). 18650 cells, like all batteries, will age with use but as Turble notes they should be replaced together (at least in banks or blocks), and one significant role of the battery management system is to mediate charge and discharge across the entire pack to maximize life for a given operating condition. Making the cells individually replaceable is not only mechanically and electrically very complicated and would significantly reduce reliability, but is also not even desirable. This is a solution to an essentially non-existent problem.