If there really is electromagnetic interference (EMI) from the power cables, then I would suspect that there is something awry with the copper 100baseT cabling (twisted pair?) such as stripping of insulation and other cable problems related to poor construction. They shouldn't normally be affected even by power cables close by.
However, you are quite right to recommend sufficient separation between data and power cables. That's what standards are for. I don't know of a means for measuring the EM interference directly – only the symptoms.
The effect on data lines should be perceived as "media errors" – this characterization is in terms of corruptions (and consequently packet loss or retransmits) on the wire. You should be able to inspect the statistics on the interfaces of the devices connected to the suspect wires and see unfavorable levels of loss.
Typically such loss is stochastic (randomly generated) and would be present at all times (assuming power is flowing through the power cables at the same time). In these respects, it should look like wireless. So you can try flood pinging (with mping at MTU packet size, for example) and checking for packet loss and/or (if retransmits are taking place instead) very high jitter on links you know are not otherwise busy.
Alternately, I developed a diagnostics system that is capable of identifying media errors (distinct from congestion for example) in terms of packet loss. I would be happy to help you to do that if you are unable to determine the problem otherwise.
I suspect that hardware tools like the Flukemeter should also be able to detect the symptoms (although I don't know if it actually identifies media errors as the problem).
This was first published in November 2004