Red Paper
International Journal of Research in Advanced Electronics Engineering
  • Printed Journal
  • Refereed Journal
  • Peer Reviewed Journal

P-ISSN: 2708-4558, E-ISSN: 2708-4566
Peer Reviewed Journal

International Journal of Research in Advanced Electronics Engineering


2025, Vol. 6, Issue 2, Part A
Model Predictive Control (MPC) for nonlinear industrial processes using machine learning algorithms


Author(s): Nguyen Minh Tu

Abstract: The study explores the integration of Machine Learning (ML) algorithms into Model Predictive Control (MPC) (MPC) frameworks for improving the regulation of nonlinear industrial processes. Traditional MPC, while effective in handling multivariable systems and process constraints, faces limitations when applied to highly nonlinear and time-varying systems due to model inaccuracies and computational complexity. The research addresses these challenges by embedding data-driven predictive models specifically Gaussian Process (GP), Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM), and Input Convex Neural Network (ICNN) architectures into Nonlinear Model Predictive Control (MPC) (NMPC) structures. Experimental and simulation results across benchmark systems such as continuous stirred-tank reactors and distillation columns demonstrate that ML-based MPC significantly reduces prediction and tracking errors, shortens settling times, and minimizes constraint violations while maintaining computational feasibility. The comparative evaluation shows that GP-MPC achieves superior prediction accuracy through probabilistic learning, RNN/LSTM-MPC balances precision with real-time performance, and ICNN/Koopman-MPC offers near-explicit optimization with minimal latency. Statistical analysis of performance metrics, including mean squared error (MSE), integral absolute error (IAE), and computational latency, validates the hypothesis that constrained, regularized ML models embedded in MPC frameworks can provide stable and efficient closed-loop control. The study concludes that ML-augmented MPC is a viable solution for intelligent automation in complex industrial environments, paving the way for adaptive, self-learning control systems capable of continuous optimization. Practical recommendations emphasize the gradual industrial adoption of hybrid ML-MPC architectures, investment in data infrastructure for model retraining, and development of uncertainty-aware control mechanisms to ensure robustness and interpretability in real-time operations.

Pages: 38-43 | Views: 6 | Downloads: 2

Download Full Article: Click Here

International Journal of Research in Advanced Electronics Engineering
How to cite this article:
Nguyen Minh Tu. Model Predictive Control (MPC) for nonlinear industrial processes using machine learning algorithms. Int J Res Adv Electron Eng 2025;6(2):38-43.
International Journal of Research in Advanced Electronics Engineering
Call for book chapter