The training runs / calibration curve will be different than with metallic pipes (and copper will be different than iron, for that matter
- and clean iron or copper will be different than iron or copper with a heavy coating of scale/sludge built up inside.) The thermal time constant will be different. But it should work just fine. Plastic pipes are far from perfect insulators.
The temperature differences don't need to be large. Effectively the approach is trying to measure how good of a heat sink the pipe is, and a pipe with flowing water is a much better heat sink than a pipe with water sitting still - even if it's a plastic pipe.
Easy, but "invasive". Valve, pressure sensor. Quite commonly done in the non-electronic regime (shut off the main valve and look at a gauge while "no water is being used.")