RR_NDB
1) You consider the possibility of the cell voltages varying outside "safe limits" during transients that as you mentioned occur starting APU?
This might be more of a steady state phenomena. This APU is an AC device, driven by a variable frequency drive (an three phase inverter). Although there is some input filtering on such units, they do draw continuous ripple current on top of the DC supply. And this ripple current will change in magnitude and frequency as the APU starter accellerates.
2) The equivalent LCR of the battery could resonate in this frequency range? (hundreds of KHz)
It is possible. Some analysis and tests could be done to characterize the batteries' AC response. An LCR circuit might be an over simplification of what is actually going on. For the real RF engineers, it could be more like a lossy transmission line, with standing waves along the battery ribbon and the resulting higher voltages at these nodes. But for simple understanding, the lumped model (as discrete components) will suffice. Here is some more info:
LC circuit - Wikipedia, the free encyclopedia
Two damage modes might be possible: First, the voltage across a series LC circuit at resonance is at a peak across the L and C (but out of phase). The C here is the capacitor created by the battery plates and seperator. If this rises above the designed insulation level of that insulator (3.7 V plus some margin of error), it could punch through and initiate a short in the cell. The other possible mechanism is electrolyte heating due to losses caused by the r.f. voltages. Some capacitors are noted for having high internal losses (dissipation factor for the EEs) and, as a result heat up (and sometimes explode
) when driven with a ripple current/frequency over their rating. Although the total voltages don't exceed the insulation's capability, hot spots can be generated.
3) The relay/contactor could generate cell voltage unsafe transients? (opening under high load)
4) The diode module could be an extra factor? (e.g. to cell "integrity" during transients)
These are possible as well, creating some high voltages due to the "inductive kick" experienced when switching high currents.
This is all guessing on my part. But given the design and certification program of the 787, it is likely that the consequences of load interaction with the battery were not completely tested. When the battery system was being developed, no actual loads (with their AC rippe) were available yet to characterize their effects. Hence the planned flight testing program.
The fortunate thing about these effects is that; should they prove to be the root cause of failres, there are some simple and inexpensive filtering techniques that can be adapted to the battery/load circuits to mitigate them.