I was assuming that there are no phase markings anywhere on the system. If there are, then it’s a lot simpler: measure the voltage between the unknown phase and the three known ones (or two known ones should be enough) - when you measure 400 - 440 V, it is not that one. When you measure a very low (< 20 V, and most likely a lot less than that, but a small voltage), you’ve got the same phase.
My method using the 3-phase sketch maybe wasn’t too clear. Assuming there are no phase markings anywhere, you start off on the downstream side of where the PV joins, probably upstream of the consumer unit, but somewhere where you have access to the three phase line cables separately.
- Calibrate the voltage. Any socket is good for this.
- Put all 3 c.t’s on the Phase 1 cable - the one with the a.c. adapter on it (by definition it is phase 1 for our purposes) and calibrate the current so that all three read the same correct current. Any substantial load will do for this.
- Take all other loads off the system. Put a pure resistive load on phase 1 and set PHASECAL for input 1 to give a real power equal to the apparent power, and a power factor of 1.0
- Set PHASECAL for inputs 2 & 3 to give a real power equal to half the apparent power and negative, and a power factor of -0.5
That should have set up and calibrated the emonTx.
With all loads off the system still:
Put a pure resistive load on phase 1 and note the power.
Transfer the c.t’s for inputs 2 & 3 to the other two phases.
Transfer the load to one of the other two phases. If either input 2 or input 3 gives the correct power, you have the c.t’s on the right cables. If not, swap the c.t’s 2 & 3, when one of them should give the right power. If it does, you have identified the correct phase rotation, so label it. If it doesn’t, something has gone badly wrong.
Then you can go back to the inverter output and identify the phases there by measuring the voltages between there and a place where you know which phase is which.