Crazy experiment idea : battery module bypass

I have two 'spare' batteries for my Gen 2 Prius. I still use the one that came with my car, but I bought a battery for $400 (in US$) from a doctor; it is a total pile of junk with all the cells imbalanced and with many cells that fluctuate, and another that I bought for $100 from a person who had a new battery installed from Toyota. The $100 battery is amazingly balanced, much better than my current battery, but one of the 1.2v cells in a module is totally dead, meaning that the module-pair has voltage fluctuations as it 'charges' and 'discharges' super quickly - the dead 1.2v cell charges in a few seconds and discharges in a few seconds under load, triggering big red triangles and so forth.

The Prius uses a HV-HV DC convertor, to give a stable voltage to MG1 and MG2 etc. The Prius C has the same engine as the Gen 2 Prius (the Prius C is 3rd Gen, but still 1.5 litre, yet the 3rd Gen is 1.8 litre). The Prius C has a smaller battery pack, specifically less modules than a Gen 2 battery pack.

In theory, given that the Prius steps the voltage from the HV battery up using a DC-DC HV convertor, it should be possible to use a battery with fewer modules than standard. Particularly as the voltage can fluctuate significantly anyway (e.g. 14.4-19 volts on my car when on very low charge versus a long period of regen and high SoC).

Does anyone know how the programmers coded their code? I don't know.

If I was to take a very good condition battery with a single very dead module-pair, and bypass that module pair, then the total (serial) voltage would be about 14.4 volts lower. This should be fine. However, the battery ECU monitors all the module-pair voltages individually. The question becomes 'did the programmers add code that will throw a code if the total voltage of all module-pairs is different to the voltage detected across the whole battery'.

- Depending on how they programmed the ECU, there is a chance that bypassing the 'dead' module pair could allow an otherwise good battery to function well.
- Risks would include major errors if the battery ECU complains that the expected voltage (by adding all module-pair voltages) is different to the voltage measured across the whole battery.
- Another risk would be if the ECUs end up overcharging the battery, so that the remaining good modules are all overcharged a bit, shortening their lifespan.

I don't have the source code, and my ECU emulator is far from complete.

I plan to bypass a module-pair, and see what happens if I connect such a battery to the car. The bypassed module-pair is an issue in that it would read 'zero volts' if the voltage sense wires are disconnected, or would read a 'stable' voltage if the voltage sense wires remain connected. I would simply connect the voltage sense wires to an adjacent cell. The 1000 ohm resistor would prevent any fires if a 'short circuit' occurs across the sense wires.

I have a few spare cars as backups if this goes wrong. I would be interested to know if anyone has any ideas what the outcome of this experiment will be? I am hoping that the car might be able to function normally. If it does, then any 'dead' HV battery that is 'dead' because a single cell in a module / module-pair has failed, might be able to be resurrected by bypassing the dead module......

 

These are the things I am wondering! The battery for the experiment is currently sitting in the back of my Leaf, and if not too tired I might just do this at the weekend!

My plan would be to take the sensor from the 'dead' module pair and use it to measure the module pair next to it (given this battery is very well balanced except for the first pair / block). This would then let the ECU think that the disconnected module is still running at the same voltage as all the other modules / blocks in the pack.

Doing some testing today, I ran the battery down to two bars, then did full acceleration uphill; I managed to briefly get 190 volts from the battery; I know that when going down a long hill and not using 'B' mode properly, the battery can temporarily reach 19 volts per block, for a total of 266 volts. This means that the HV-HV convertor can actually run from a wide range of voltages, so losing approx 14.4 volts from one pair / block should not make a difference to the actual function of the car.

If the ECU code measures the individual cells and deltas, and the disconnected cell is 'mirrored' (on the voltage sense lines) to a good cell, then the ECU should (when comparing the cells to each other), be happy.

If the code that manages the HV-HV convertor is somewhere else, for example the HSD / Inverter ECU, then there is a chance that it will read the voltage of the battery, and then convert that voltage to whatever is needed for the HV system / MG1 / MG2 etc.

The issue then becomes; does the battery ECU run a check where it adds up each of the blocks, to make a total, and then compares this to the total voltage across the battery? If so, it will detect a fault. The issue then becomes how it would interpret such a fault; it might be possible that the car would be perfectly usable, or not.
- It seems logical that such code exists. However, I am not sure how fast the sample rate is as the ECU cycles through the blocks. Technically it could be possible during sudden braking (and thus rising voltages), the calculated voltage (each block added together) could be quite different to the 'total' battery voltage. This kind of issue would require that the programmer includes a certain degree of tolerance. Given the possible battery voltage range of 190-266 volts or more, it is not certain that the experiment will be a failure. However, I would still be very surprised if the car works fine with such a modification!

 

True; however, the MGs get their AC from the DC battery (also, via all kinds of other ways such as the other MG etc). However, one example would be; in 'EV' mode, doing max acceleration, and also doing downhill braking (max regen at say 44km/h), the voltage of the battery changes a lot. During the acceleration, the MGs are getting their 3-phase from a fluctuating voltage source (the battery).

The Prius C uses a lower-voltage HV battery with fairly similar electronics otherwise; I am not sure if the car is necessarily 'hard-wired' to a certain number of modules; the NWH20 battery ECU itself is made for battery packs from the NWH11 anyway (except for the fact that the board is not populated with the components at the HV side, but it was clearly designed for a larger pack). The software will be hard-coded, almost for certain, though!

 

Me too; the question would be : how does the car decide what to charge the battery to? It could be based on the individual block voltages, or it could be on the overall voltage?

Either way, the battery ECU does stop charging when even a single block goes too high / and vice versa when a single block goes too low (or if they all go low). It all depends on exactly how they programmed the software.

What is interesting is that during periods of full-regen, the per-block voltage can almost reach 20 volts; while the cells aren't really 20 volts (just the chemistry 'settling down' after a sudden burst of charge), it shows that, in a way, the 'capacity' can't really be determined by the ECUs by either total, or module / block voltage.

It likely uses the current gauge to 'guess' where it is, and if it sees the cells reaching high voltages for too long, it reduces the max regen and uses the friction brakes more.

The only way to test would be to actually do the bypass; I have so little time but hope to do it soon. It really depends on the software algorithms. if it was possible to hack the battery ECU (which it isn't, yet), it should be simple to reduce the block count down to 13.