(409a) Distributed Control System Design Using Lyapunov-Based Model Predictive Control

Modern chemical plants involve highly integrated process units and utilize automatic control systems to meet desired safety, environmental and production specifications. Most control systems are decentralized in nature from a plant point of view, that is, each control system operates on a process using local process information without communicating with control systems operating on adjacent units. Nowadays, with the new developments in communication technologies such as wireless networks, there is an increasing interest in developing distributed control architectures; that is, the various control systems share information in order to take their corresponding decisions autonomously. This distributed control architecture gives rise to many interesting control problems. On one hand, the shared network used by each control system may introduce new dynamics into the closed-loop system such as time-varying delays or data losses. On the other hand, the distributed control algorithms must take into account how local control decisions affect the other control systems and affect the overall closed-loop performance. This is in general a very difficult problem. There are many works in the literature that address distributed control and its applications to different fields such as multi-agents control, sensor networks or internet congestion control, and more recently, chemical processes.

In this work, we introduce a distributed control architecture based on Lyapunov- based Model Predictive Control (LMPC) for nonlinear process systems with both continuous and asynchronous sensing and/or actuation. This class of nonlinear systems arises naturally in the context of process control systems based on hybrid communication networks (i.e, point-to-point wired links integrated with networked wired/wireless communication) and utilizing heterogeneous, asynchronous measurements. The control problem is to coordinate the design of a local (single-process) Lyapunov-based model predictive controller which takes decisions based on synchronous process sensing and actuation, and a supervisory (plantwide) Lyapunov-based model predictive controller that utilizes networked sensors and actuators and accounts for asynchronous sensing, actuation and network behavior, to ensure closed-loop stability and performance. To address this problem, when a new asynchronous measurement arrive, the supervisory controller solves a plantwide MPC optimization problem assuming that the local controller implements an already existing Lyapunov-based control law that guarantees closed-loop stability. The optimal input trajectory is sent to the local controller along a contractive constraint in the form on an upper bound on the value of the Lyapunov function along the predicted closed-loop nominal state trajectory. This contractive constraint is utilized in the MPC optimization problem of the local controller and guarantees that even in the local controller does not bahave as the supervisory controller assumed, stability of the closed-loop system is preserved. The proposed distributed control architecture profits from both the continuous and the asynchronous measurements as well as from additional networked control actuators and preserves the stability properties of the local controller even in the presence of asynchronous communications while improving the closed-loop performance. The theoretical results are demonstrated using chemical process examples.