Scenario-based approaches for handling uncertainty in MPC for power system frequency control

Abstract

A stochastic nonlinear model predictive controller (SNMPC) is designed for automatic generator control of a proxy of the Nordic power system, and it is compared with a multi-stage nonlinear model predictive controller (MNMPC). Both controllers are scenario based, but originate in two different disturbance modeling paradigms; stochastic and deterministic. A simulation study indicates that the two controllers behave similarly. The MNMPC is however less exposed to infeasibility issues, and it also has better tractability than the SNMPC. On the other hand, the SNMPC gives probabilistic guarantees for constraint fulfillment; a feature whose practical implications are debatable.