If U1 is a standard op amp, the circuit in the OP will be operating as a comparator indicating the polarity of phase A vs neutral, not indicating the relative voltage, because the op amp has no feedback. I would guess it's meant to be an integrated differential amplifier, in which case the circuit should work fine (with proper resistor values) for phase-to-phase as well as phase-to-neutral measurements. R1/R2/R5 and R3/R4/R6 scale the Phase A and Neutral voltages respectively relative to VREF (presumably a mid-rail reference to handle positive and negative input voltages), then U1 amplifies the difference in the scaled voltages. The circuit is entirely symmetrical, other than the sign of the output, so it doesn't matter which way around it's connected.
Alternatively, the reference point for any circuit is arbitrary, so you could use one of the phases as the reference for all of the voltage measurements. The measurement circuit can be directly referenced to that phase, with voltage dividers to sense the voltages on the other phases with respect to the reference phase. The measured results can be sent over an isolated digital link to ground-referenced circuitry if needed -- there are even specialized energy management ADCs designed for that. Of course it's a bit harder to test and troubleshoot a circuit that is directly referenced to a phase, but it might be worth the effort in some cases.