Control strategies for low frequency disturbance rejection

We search linear feedback law for suppressing a disturbance, in particular a strongly autocorrelated disturbance process. Before augmenting control system with feedforward, we want to investigate achievable performance with feedback only. PI design clearly yields limitations: Disturbance effect could not be suppressed over all frequencies. What alternative feedback controller could achieve good disturbance rejection?

Note: We do not have latest Control system toolbox. We know currently it is possible to have Matlab calculate controller for a specified sensitivity profile. However we look for 'theoretical hints' as to what controller structure should give better disturbance suppression than a simple PI control.

Answer 1

[Kwin]

Depending on whether you are dealing with input or output disturbance, then you either want to minimize the process sensitivity (H/(1+CH)) or just the sensitivity (1/(1+CH)) respectively. In case of the sensitivity then you will always have that after a certain frequency you have that its magnitude will be close to one (also you will have to deal with the waterbed effect). All you can do, assuming that the closed loop system says stable, is increase the crossover frequency (the frequency near which the magnitude of the sensitivity gets close to one) and suppress a as high a frequencies as possible. If this causes stability margin issues, you could maybe add a lag-filter (which helps bending away from the -1 point in the Nyquist diagram).