Normally two PIDs are used to do torque-control with FOC -- one for the direct torque and one for quadrature torque.
The reason for this is described here:
Assume that I characterised a particular motor so that for any given speed and torque (as measured by current and speed sensors) the ideal phase-offset was known (or interpolated) by a lookup table.
If I used this lookup table combined with sensors for rotor-position, torque, and speed, would I be able to dispense with the second PID loop and only use one for determining the quadrature (torque) force?
I believe this question is equivalent to asking whether the phase-offset depends on anything other than the speed and measured current. If there's no other forces that might affect the phase-angle, then the PID for direct torque can be removed.