Ask Greg McMillan

Greg's Response:

There are many measured and unmeasured disturbances to process inputs (load disturbances). A feedforward correction needs to arrive at the same point in the process at the same time as the load disturbance with an effect that is equal to and opposite of the load disturbance. Dynamic simulation is essential to finding load disturbances and identifying feedforwards, and the dynamic compensation to provide the corrections with the right timing and size. The compensation changes with operating conditions and may need to be suppressed if unmeasured disturbances are driving the controlled variable back to setpoint. If the feedforward arrives too soon, inverse response can occur. If the feedforward correction is too large, a disturbance in the opposite direction is created. These situations cause oscillations that are confusing to the process PID controller and operations. 

Dynamic compensation of the feedforward signal by means of a lead-lag block and a gain are used to provide the correction with the right timing and size. The feedforward lag is set equal to the load lag (disturbance variable path lag), the feedforward lead is set equal to the loop lag (controller output path lag). The feedforward gain is the loop gain divided by the load gain and is generally set slightly lower to prevent creating a disturbance in the opposite direction due to nonlinearities and unknowns. Dynamic compensation can be adapted based on how loop and load dynamics change with operating conditions. Intelligence can be added that detects how much an unmeasured disturbance is causing the controlled variable to move in the same direction as the feedforward correction requiring suppression of the feedforward correction. 

Dynamic simulation with a Future-Value block can predict how much an unmeasured disturbance will drive the controlled variable back to setpoint. The feedforward gain could then be reduced by the ratio of the unmeasured load predicted error to the measured load predicted error. An adaptive PID controller can be setup to adjust the feedforward gain to minimize the feedback correction of the feedforward done by the process controller. Simulation can find the adaptive PID tuning to provide a gradual approach to zero feedback correction of feedforward by the process PID that minimizes interaction between the process and adaptive PID controllers.