Semi-infinite programming yields optimal disturbance model for offset-free nonlinear model predictive control

Offset-free nonlinear model predictive control (NMPC) can eliminate the tracking offset associated with presence of plant-model mismatch or other persistent disturbances by augmenting plant and employing an observer to estimate both states disturbances. Despite their importance, a systematic approach for generation suitable disturbance models is not available. We propose optimization-based method generate based on sufficient observability conditions generalize theory offset-free NMPC allowing (i) more measured variables than controlled (ii) unmeasured variables. Based conditions, we formulate generalized semi-infinite program, which reformulate solve as simpler program using discretization algorithm. The solution furnishes optimal model, maximizes set those state, manipulated variable, realizations, condition satisfied. generated offline be used online NMPC. apply three case studies ranging from small scale chemical reactor cases medium polymerization case. results demonstrate validity usefulness show that successfully finds