کنفرانس مهندسی برق ایران

صفحه اصلی / سی و دومین کنفرانس بین المللی مهندسی برق

Stochastic model predictive control based on online learning for a class of nonlinear constrained systems

نویسندگان :

Mahdi Mansoury¹ Mohammad Ali Badamchizadeh² Hamed Kharrati³

1- دانشگاه تبریز 2- دانشگاه تبریز 3- دانشگاه تبریز

کلمات کلیدی :

Gaussian process،Nonlinear systems،SMPC

چکیده :

The progress in machine learning and advancements in measurement and computational capabilities in modern wave control systems has spurred interest in learning-based control techniques, including model predictive control. The primary challenge in model predictive control lies in addressing the inherent uncertainty in complex and dynamic system models. This paper explores the effectiveness of online system model learning to use Gaussian processes. Recent years have witnessed significant results by combining non-parametric learning models with Gaussian processes and model-based predictive control. This research focuses on assessing the effectiveness of Gaussian processes, considering a stochastic model predictive controller (SMPC). To evaluate the proposed method, simulations on a quadruple tank process are conducted, aiming to regulate liquid levels. Gaussian processes are utilized for real-time estimation of system hyperparameters, incorporated into the predictive control cost function, and account for input and output constraints. The results show the stability and efficacy of the proposed control system.