How do you design feedforward control and when is it beneficial?

Feedforward measures disturbances and acts before they affect the controlled variable. Design: determine disturbance-to-output and manipulated-to-output transfer functions, feedforward controller = -Gd/Gp (inverse of process model). Benefits: immediate response to measured disturbances (faster than feedback), can achieve perfect compensation theoretically. Limitations: requires accurate process model, only handles measured disturbances, errors accumulate without feedback. Usually combined with feedback: feedforward handles measured disturbances, feedback corrects unmeasured disturbances and model errors.