Oh you actually want this explained - I didn't understand that from the earlier responses.coleafrado wrote: ↑Sat Jun 20, 2020 7:44 amSo you're not going to back this up with anything?
This is not very obvious to many people because there is some misleading 'marketing' (or whatever you can call it) around highly-parallellized packs. The idea most people have about Tesla packs is that even if one cell fails, there are many in parallel that can take up the slack. This is an oversimplification and in general is simply not the case, because:
- The remaining parallel cells will have proportionally higher load on them, during discharge (generally fine) but also during charging. Even a few percent higher charging current will impact charging rates at high SOC quite significantly. Again, see the Leaf batteries for what happens when a pack anisotropically ages - you get disproportionally rapid degradation and imbalance
- The entire rest of the pack is reduced in capacity proportional to just that one failed cell.
- That cell didn't fail for no reason; there is almost never a single incidental cell or bonding failure, it's always a cluster affecting either multiple parallel cells or single adjacent series cells as well.
- You're introducing - in the case of a full Tesla pack for instance - roughly 10k-15k points of failure per battery pack.
- Cooling of cylindrical cells is already quite hard, but especially when there's a failed and possibly deformed or venting cell nearby.
Within the industry it's pretty well known what the Tesla battery failure rate is. You won't be surprised that especially near the beginning of manufacture, it's rare to find packs without bonding faults, even in otherwise healthy packs. Within the battery refurbishment industry which we rely on heavily, it's basically impossible to refurbish entire Tesla packs. There is always one or two S/X modules that have a serious issue in replaced packs that enter the waste stream, and most of this has to do with their packaging strategy.
Now, it's also pretty well known in the industry that all of this has become much, MUCH better over the years. They've got 7 years of wire bonding experience to perfect the tech, but that doesn't mean wire bonding as a general technology has progressed - it's just Tesla that has improved by and large. Since last year - i.e. before we started mass production of our extender batteries - we have been looking around for alternatives to the batteries we're using now (large prismatics in modules with BMS attached), and one of the cheaper options would have been to buy a wire bonding machine and enter a supply contract with Samsung SDI or LG for 18650s. But we're getting quoted - at best - 0.1% bonding faults. Even premade packs have these kinds of reliability ratings. This is why wirebonded packs are basically never sold for automotive uses.
A second line of critique on using 18650s in a non-thermally managed pack is their poor performance. Intrinsically, 18650s have bad packing density, i.e. you spend a lot of volume and weight on packaging compared to active material when compared to prismatics. To compensate, they have to either increase their active-material-to-electrode volume or decrease their energy density. This means for most 18650s, their specific power is quite low, at least for commercial offerings you can consider for EVs.
An NCR18650B, the 'tesla cell', has a continuous discharge rating of 3.35A and a peak of 6.7A. You can fit, realistically, something like 40kWh worth of these cells inside a Leaf battery enclosure, or about 96S40P. That translates into a 140/280A continuous/peak rating, or about 50/100kW continuous/peak - at 25C. The 24kWh pack of yesteryear has a cont/peak rating of 30/80kW at freezing! So you're looking at a significantly lower performance, especially at low temperatures.
Tesla gets around this by - first - just making giant battery packs (more parallel solves all these problems), but also making sure the pack never operates at low temperatures to begin with and gets cooled down really aggressively at high performance. You need to remove roughly 10kW of heat when charging at (old) supercharger rates with a 40kWh pack made out of NCR18650Bs, or about 2.5kW at chademo rates. Again, you can compensate for a lot of bad design options by improving in other ways. Tesla chose a shit battery construction from a reliability, maintainability and performance point of view, but coupled it to absolutely excellent thermal management to make batteries that can do anything you'd want from an EV battery and more. Better than a lot of their competition even, which is a truly underappreciated accomplishment.
Additionally, 18650s have bad overloading and aging characteristics. Again, bit less of an issue with packs that are well-managed, but lithium ion cells expand not just when they get (over)charged, but when they age as well - building up pressure inside the mechanically constrained shell. At about 70-75% SOH, a cell expands 5-10%, and that's something you can account for in prismatics - but not in 18650s. They just fail - and if you have thousands in a pack, that's a lot of opportunities for failure again.
All of that is why I am VERY critical of 18650-based rebuilt batteries. It is a good choice for a multi-billion dollar company whose core technology is these batteries. It is a bad choice for a 2-man startup without a rigorous background in battery engineering. There is a lot you can do wrong with battery design in general, and all that gets magnified if you use cells that can barely handle the load, and need specialized BMSes to survive for a decent amount of time - oh and there's tons of labor going into production and QA, and it's mighty attractive to skimp on that last bit to preserve your margins.