(522g) Control Lyapunov-Barrier Function-Based Predictive Control of Nonlinear Processes Using Real-Time Machine Learning Modeling | AIChE

(522g) Control Lyapunov-Barrier Function-Based Predictive Control of Nonlinear Processes Using Real-Time Machine Learning Modeling

Authors 

Wu, Z. - Presenter, University of California Los Angeles
Rincon, D., University of California, Los Angeles
Christofides, P., University of California, Los Angeles
Safety considerations have been integrated into feedback control system design to yield cooperative actions to ensure both operational safety and economic performance of system operation via control Lyapunov-barrier functions (CLBF) [1,2]. Specifically, in [2] a CLBF-based model predictive control scheme was proposed to optimize process performance accounting for process stability and safety requirements. While MPC systems implemented in an industrial setting often utilize linear data-based empirical models to compute control actions to maintain optimal process operation while accounting for process and control actuator constraints, chemical processes are inherently nonlinear and often require nonlinear models in order to be controlled efficiently. Modern machine learning modeling tools provide an efficient way to nonlinear process modeling, taking advantage of advances in training, efficient coding and computational power [3, 4, 5].

In this work, we develop a machine-learning-based CLBF-MPC based on an ensemble of recurrent neural network (RNN) models that are widely-used to model nonlinear dynamic systems for prediction to control an input-constrained nonlinear process accounting for stability and safety considerations. Specifically, RNN models are first developed to model a general class of nonlinear systems using process operating data, and sufficient conditions that account for bounded modeling error between the RNN model and the actual nonlinear process are provided to achieve closed-loop stability and safety for the nonlinear process under CLBF-MPC. Additionally, following the design of machine-learning-based CLBF-MPC, the CLBF-based economic MPC using RNN models is proposed to optimize process economic benefits as well. Moreover, to handle model uncertainty issue in real-time implementation of controllers, on-line learning of RNN models is also employed within CLBF-MPC and CLBF-EMPC to update process models in the presence of time-varying disturbances.

[1] Romdlony, M. Z., and Jayawardhana, B. Stabilization with guaranteed safety using control Lyapunov–barrier function. Automatica, 66, 39-47, 2016.

[2] Wu, Z., F. Albalawi, Z. Zhang, J. Zhang, H. Durand and P. D. Christofides, "Control Lyapunov-Barrier Function-Based Model Predictive Control of Nonlinear Systems,'' Automatica, 109, 108508, 2019.

[3] Kosmatopoulos, E. B., Polycarpou, M. M., Christodoulou, M. A., and Ioannou, P. A. High-order neural network structures for identification of dynamical systems. IEEE transactions on Neural Networks, 6, 422-431, 1995

[4] Mohanty, S. Artificial neural network based system identification and model predictive control of a flotation column. J. Process Control, 2009, 19, 991−999.

[5] Wu, Z., A. Tran, D. Rincon and P. D. Christofides, "Machine Learning-Based Predictive Control of Nonlinear Processes. Part I: Theory,'' AIChE J., 65, e16729, 2019.