Algorithms and software tools for application of MPC in process control systems

Abstract

This workshop is focused on development and usage of algorithms and software tools for easy and effective application of Model Predictive Control (MPC) in process
industries. Specifically, we present two open-source packages developed in Python language. The first package is a comprehensive modeling and simulation software for linear and nonlinear MPC systems. It offers a high-degree of flexibility in terms of objective function type (linear/quadratic/nonlinear, tracking vs. economic, continuous-time vs. discrete-time), state and disturbance estimator (static observer, Kalman filters, Moving Horizon Estimators), nominal vs. offset-free design, different type of constraints (input, states, outputs, mixed). The second package is a system identification software that includes different linear system models (ARX, ARMAX, state-space) and identification methods (prediction error and subspace identification). The package allows automatic model order selection or user choice of model orders, and is intended to be used with minimal knowledge of system identification principles. During the workshop we analyze a number of examples taken from process control applications. Finally, we also present a variant of the two packages that allows to model the presence of valve stiction (static friction), and we analyze recent MPC algorithms that can effectively cope with such problem, commonly encountered in process industries.