Stochastic model predictive control based on multi-step control strategy for discrete nonlinear systems

Stochastic model predictive control (SMPC) is a popular approach to control uncertain systems by incorporating the probabilistic distribution of the uncertainty into the controller design. However, the performance of the SMPC deteriorates rapidly when the system becomes nonlinear. In this article, an efficient SMPC is derived after decomposing the discrete nonlinear systems into a linear stochastic model and a bounded model mismatch. The proposed controller includes a linearized-model-based SMPC and an RMPC for the model mismatch. The multi-step control strategy is incorporated to reduce the conservativeness by introducing more degrees of freedom. The feasibility of the recursiveness of the proposed controller is thoroughly discussed. The control performance is illustrated by examples involving both a numerical nonlinear stochastic system and a wind-blade pitch system.

Automated summary

This article presents an efficient stochastic model predictive control method that decomposes discrete nonlinear systems into a linear stochastic model and a bounded model mismatch. By incorporating a multi-step control strategy, the conservativeness of the controller is reduced. The feasibility of the recursive nature of the proposed controller is thoroughly discussed, and the control performance is demonstrated through examples involving numerical nonlinear stochastic systems and a wind-blade pitch system.