Good advice dpc.
I usually try to direct connect as much capacity as I can without overvoltage problems at light loads.
I also start by calculating how many KVARHrs I need to correct the power factor to 90% (Or whatever figure your penalty kicks in at).
I divide this by the number of hours in a month to get my needed KVAR rating. It is often possible to avoid the expense of a power factor controller.
The resulting KVAR rating of the capacitor bank is much smaller than the KVAR capacity required to provide 100% correction under all circumstances.
The power company is penalizing you for using more KVARHrs in a month than you are entitled to based on your KW consumption.
Power factor is an easy way of expressing this relationship, but at the end of the month, when you write the check for the power consumption and any penalties, The bottom line is "You used too many VARHrs."
I realize that there are regional exceptions to this general statement. I understand that some utilities may use either mechanical or software ratchets on the meters to prevent subtracting VARHrs when your power factor goes leading.
My general comments are still applicable.
If, after evaluating the system, it is determined that more capacity is needed than can be safely permanently connected, look for motors that run 24/7. A good example is the fans in a lumber drying kiln at a sawmill. If all the kiln motors are corrected to 100% PF, this often goes a long way to bringing the plant power factor up to the penalty cut-off point. The kiln fans are often started in groups. It may be more economic to use a signal from the kiln control panel to bring in one large bank of capacitors when the motors have started rather than correcting each motor individually.
I realise that every-one doesn't have a sawmill to use for power factor correction, the point is, look for the largest loads you can and see what you can do with them.
In the case in question, I would start by installing enough primary capacitors to correct all the transformers to at least 100% power factor, and see how many KVARHrs I can get and how close to the penalty cut-off I am. If I need more I would consider enough capacitors to over correct the transformers to 90% or 80% leading.
If you need still more capacity, look around for a big motor that typically runs for most of the operating hours of the facility. Step one, correct it to 100% power factor. If you need more capacity look for similar motors to correct or use a contactor to connect a capacitor bank directly to the line that is about twice the capacity required to correct the motor to 100%. Typically, if the large motor is running there will also be a number of small motors running at the same time to absorb the extra KVARs.
Power factor correction can be a science or an art.
The science is to specify a power factor correction unit.
If your budget is less forgiving than Haliburtons, you can save a lot of expense by resorting to the "Art" of power factor correction.
Look around, use ingenuity, and you can often drastically reduce the cost of power factor correction.
Re transformer correction. Measure the no load current of the transformer. Select a capacitor bank that will result in an equal current and you are close enough to 100% correction. Very much closer than the 15% tolerance allowed in the ratings of power factor correction capacitors.
Respectfully.