Stochastic Nonlinear Model Predictive Control with State Estimation by Incorporation of the Unscented Kalman Filter

Abstract

Nonlinear model predictive control has become a popular approach to deal with highly nonlinear and unsteady state systems, the performance of which can however deteriorate due to unaccounted uncertainties. Model predictive control is commonly used with states from a state estimator in place of the exact states without consideration of the error. In this paper an approach is proposed by incorporating the unscented Kalman filter into the NMPC problem, which propagates uncertainty introduced from both the state estimate and additive noise from disturbances forward in time. The feasibility is maintained through probabilistic constraints based on the Gaussian approximations of the state distributions. The concept of robust horizon is introduced to limit the open loop covariances, which otherwise grow too large and lead to conservativeness and infeasibility of the MPC problem. The effectiveness of the approach was tested on a challenging semi batch reactor case study with an economic objective.

6 pages, 5 figures

References in corpus (1)

• Stochastic Nonlinear Model Predictive Control with Efficient Sample Approximation of Chance Constraints

Cited by in corpus (4)

• Stochastic data-driven model predictive control using Gaussian processes
• Hybrid Gaussian Process Modeling Applied to Economic Stochastic Model Predictive Control of Batch Processes
• Combining Gaussian processes and polynomial chaos expansions for stochastic nonlinear model predictive control
• A software framework for stochastic model predictive control of nonlinear continuous-time systems (GRAMPC-S)