First of all, let me say that I am deeply indebted to Lee Hart, How Kuff, Brian Leeper, Mark Pullen, and Chris Edwards. Without the thoughts and insights of these people, the monitoring system would not have gotten as far as it has (not that it is by any means complete - there are still many, many rows to hoe).
Ask any group of electric-vehicle owners what the biggest pain in their sides is, and the answer you will invariably get is 'Batteries'. With the exception of a few bad designs and a few user errors, controllors are solid state beasts that pretty much should Just Work (tm), and the motors often show operating times of 50-60,000 hours in actual use in industrial facilities. (Obviously, no EV driver gets to put that many hours on them). While DC motors are also occasionally a little flaky due to the high voltages and the evils of brushes, etc, they are nowhere near the problem child that batteries are.
One thing that I keep noticing over and over is that very many times, a computer watching a few basic parameters could have stopped some of these rather explosive failures before they became serious. ANd, since I hate walking home, I decided to apply my computer knowlage to the problem.
Also, since you kind of need to be able to monitor each individual cell in order to spot failures in them, the battery monitoring system had better be cheap per-cell. And yet, you'd really like it to be pretty precise, per cell. On the surface, these things are pretty contradictory. I think I have found a solution to this problem.
Obviously, the first thing you need is a way to mate the computer to the batteries. Since they are a series string that shows a potential difference of several hundred volts, this way better include some isolation, or you're going to be listening to the snap,crackle,pop of each little IC. In addition, since you probebly aren't going to have a way to turn this off, it better be pretty damn cheap on the current-draw side of things.
Fortunately, isolation is a common problem nowadays, and there are several components readily available that solve these problems. My favorites are the isolation transformer and the optoisolater.
It's important to remember, even with isolation circuitry, that 288V DC is a LOT of juice - circuit board design must also be set up to minimize the chance of arcing, and resistors and clamping zeners should be in place in case (god forbid) the cell under a monitoring module should fail open.