(17d) State Estimation and Model Predictive Control of Nonlinear Processes Using Recurrent Neural Network Models

Machine learning techniques have demonstrated their capability in approximating complex, nonlinear and uncertain chemical processes in the discipline of data-driven modeling, and have earned attention of the research community in process modeling and control over the past few decades [1,2,3]. Among different machine learning techniques, the one that has been commonly utilized for modeling a general class of dynamical systems is the recurrent neural network (RNN). RNN-based model predictive control (MPC) that utilizes RNN models as the prediction model was developed in [3] based on the assumption that state feedback is available for data generation and online implementation of MPC. However, this assumption does not hold in many practical applications, and thus, makes it challenging to implement machine learning-based control to processes with unmeasured state variables. [4].

To address this issue, this work develops RNN-based state observer and MPC for nonlinear systems. A well-fitting RNN model is used to predict nonlinear processes dynamics. Dataset will be generated from extensive open-loop simulations within a predefined operation region to train and develop RNN model that will capture process dynamics for a general class of nonlinear systems. By using the measured states, the state observer will use the RNN model to estimate the unmeasured states. Afterward, the Lyapunov-based MPC (LMPC) that also utilizes the developed RNN models as the prediction model is developed to achieve closed-loop stability in terms of having the closed-loop system states to be bounded within the stability region for all times and eventually will converge to a small neighborhood around the origin. A chemical reactor example will be used to demonstrate the performance of the LMPC using RNN-based observer.

[1] Sontag, Eduardo D. "Neural nets as systems models and controllers." Proc. Seventh Yale Workshop on Adaptive and Learning Systems. 1992.

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

[3] 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

[4] Bequette, B.W., Nonlinear control of chemical processes: A review. Industrial & Engineering Chemistry Research, 1991. 30(7): p. 1391-1413.