You will not know the direction of AC current by measuring the current alone. You could by using the voltage and current phase relationship. It might make an interesting project to come up with a circuit which used a clip-on AC current probe and some kind of capacitance linked AC voltage probe to monitor the phase. It would require several AC current probes to figure things out and come up with a total of the power feed into the grid Vs the power being used by the house.
Its an interesting problem. You have actually three currents to consider, :- 1 current flowing from your PV inverter, 2. the current taken from the grid, 3 the current flowing to your local load. I think the sum is :- I2 = I3 - I1. The only thing that is certain is that if you are supplying power to the grid your locally generated voltage must be greater then the incoming mains voltage, if only by a few millivolts. Just thinking of numbers, domestic supply, 230V @100 A at 2% regulation (probabley worse then this!). so when 100A is drawn, the voltage drops by 2% or 4.6V. This means that the mains has an "impedance" of 46m ohms. Like wise if you want to "push" 100A into the mains you need an extra 4.6V over the nominal mains voltage. What I would do is to get two mains to 12V AC transformers, label each one, I/C and PV. Feed two bridge rectifiers each with its own resevoir capacitor and identical load resistors. Now you can measure the difference between the two voltages (~18V), with no PV it will be say +18V and with no mains and full PV it will be -18V. The range you will be interested in will be 0 -> -.36V. So feed this voltage via current limiting resistors and limiting diodes into an OP amp and you should get a nice big voltage swing indicating when you are supplying the grid.
I have been considering this problem a bit more, no real change in philosophy. One point is that the mains voltage must be monitored as close to the incoming mains as possible, not on your switchboard. I think that your PV generated voltage must be monitored as close to the inverter as possible. The reason is that you must rely on the voltage drop down the cables to separate the I/C and PV voltages. the corollary to this is measuring two points on the mains fuse box and consider them to be some how different (other then the voltage drop across your fuses and/or MCBs).
Did you ever do Kirchoffs law at College?, this is a typical problem. Take a "T" network of resistors, feed two different voltages into the upper arms, work out the voltages/currents through the various resistors. Depending on the voltage differences and the size of the resistors, the results can be very different. You will have to work out the various resistor values from your local configuration.
