Using optos for this application sucks.

1) They will put a constant load on each cell and eventually discharge them.

2) Unless you use the fancy linear type optocouplers that have the calibrating feedback path you will find the normal type couplers to be nonlinear and highly temperature sensitive.

3) Facing the fact that there are connections between the various cells in the stack the optos across each cell will be directly affected by the current paths from higher cells and those below that feed to the GND rail. This would disrupt any accuracy of reading any one cell.

Last comment I'll make is that you say that you want to not have to be concerned about the order of the cells and their location in the stack. If that is the case then why are you monitoring?? I thought the point of monitoring was to be able to isolate the cell that may have a problem.

Michael Karas