It depends, I'm afraid.

Traditionally capacitors were sited at the motor, and this still makes sense in many cases. In the language of electronics, the motor appears as an inductor in parallel with a load dependant resistor, and we are trying to tune this to parallel resonance, so the supply sees only the resistive part. Conceptually this is attractive, but it's not so simple.
Firstly the correction required is motor dependant,  and will vary somewhat with loading. Star delta starters are especially problematic, and really require two different capacitor values for the 2 states. Capacitors suddenly connected to the mains at the wrong part of the cycle can lead to larger inrush currents, already a problem with bigger motors.

Correction at the origin of the installation instead is expensive but simple and becomes more useful the more complex the installation - all the wiring needs to be rated for the VA current, not the load current, but the big box can be set up with automatic switching that brings in or removes capacitors to order to maintain some target range of power factor, usually 90-95 % . (the on board electronics looks at the current to voltage phase shift and has a binary weighted array of capacitors and relays of increasing beefiness to choose from to maintain the target phase shift.)

Actually any load can be connected to a supply with such an arrangement, as if it does not need a capacitor the box of magic won't add one...

Note that motors on VFDs are not running at mains frequency at all and do not require, and the installation may be damaged by, capacitors at the motor side of the inverter. Also all modern VFDs will have internal PFC, so presenting a near resistance to the supply and don't need supply side PFC at all either. Though local  EMC filters are often a good  idea as the waveforms are pretty horrid.

Mike.

Great explanation, Mike. I would like to add that while centrally controlled, automatic PFC systems typically offer the highest return on investment, locally installed PFC devices can also mitigate copper losses within the installation. However, it is important to recognise the limitations of individual correction methods. These methods may not consistently achieve an overall power factor of 0.95 lag or better, thereby necessitating the installation of automatic correction systems at the main incomer position.

Thanks for the reply Mike.

So would I be correct in thinking I should separate all the VFD main incoming feed from the DOL feed then have the main DOLs fed from the automatic PFC unit?

mapj1 said:

- all the wiring needs to be rated for the VA current, not the load current, but the big box can be set up with automatic switching that brings in or removes capacitors to order to maintain some target range of power factor

I don't full understand this part. The VA current will be greater but if the big box is adding and removing capacitance, why is this?

OK the wiring between motor and caps is carrying the larger current - in effect its part of the tuned circuit.

So would I be correct in thinking I should separate all the VFD main incoming feed from the DOL feed then have the main DOLs fed from the automatic PFC unit?
Not strictly necessary, as the automatic unit will not see the VFDs as needing correction, and will just correct inductive loads, but it may make the wiring in and out of the PFC unit a bit easier to manage.

Mike

OK thanks again Mike!