3 ms·

I see, that makes sense. A potentially better (and still simple) algorithm you could use is to take the differences between the cell and the lowest cell, and tr

by jroll 15y ago

I see, that makes sense. A potentially better (and still simple) algorithm you could use is to take the differences between the cell and the lowest cell, and try to balance to the lowest cell's value. This should give a slightly tighter pack, and has a side benefit of the ability to balance at any state of charge. I can send a pull request if you'd like to do some benchmarking. :) That also seems like a small voltage swing. Do you have any more information on the cells you're using? Maybe a voltage vs. SoC chart? Edit: wording

ricksta 15y ago
Actually my original algorithm did exactly that. The problem with that approach is the amount of energy required to change the cell voltage by 0.1V goes up exponentially once it's below 3.5V, where the charge curve flattens out. An example would be a 40% state of charge battery could be 3.30V and a 60% state of charge battery would be 3.33V. When the batteries are usually about 1-3% difference in SoC, it become very difficult to balance them based on just looking at the voltage alone. The measuring error is probably more than the actual voltage difference. That why I decided to not bother balancing the battery once they are below 3.5V. The numbers I'm getting is for LiFePo4. There are many other chemistry of Li-on variant, and LiFePo4 so far is the safest. It does not burn up like the ones in your laptop, but at the cost of slightly less energy density. The type of batteries we had in our EV was these ones made by Winston Battery: http://en.winston-battery.com/index.php/products/power-battery http://en.winston-battery.com/index.php/products/power-batte... They were kind enough to donate almost $20,000 worth of these babies to us. We had 160Ah x 3 in parallel.
- ricksta 15y ago
  and of course the BMS can work with any other type of Li-on battery chemistry by changing it's protection parameters.

by jroll 15y ago

ricksta 15y ago

Actually my original algorithm did exactly that. The problem with that approach is the amount of energy required to change the cell voltage by 0.1V goes up exponentially once it's below 3.5V, where the charge curve flattens out. An example would be a 40% state of charge battery could be 3.30V and a 60% state of charge battery would be 3.33V. When the batteries are usually about 1-3% difference in SoC, it become very difficult to balance them based on just looking at the voltage alone. The measuring error is probably more than the actual voltage difference. That why I decided to not bother balancing the battery once they are below 3.5V. The numbers I'm getting is for LiFePo4. There are many other chemistry of Li-on variant, and LiFePo4 so far is the safest. It does not burn up like the ones in your laptop, but at the cost of slightly less energy density. The type of batteries we had in our EV was these ones made by Winston Battery: http://en.winston-battery.com/index.php/products/power-battery http://en.winston-battery.com/index.php/products/power-batte... They were kind enough to donate almost $20,000 worth of these babies to us. We had 160Ah x 3 in parallel.

ricksta 15y ago
and of course the BMS can work with any other type of Li-on battery chemistry by changing it's protection parameters.