(687e) Quasi-Decentralized Control of Networked Process Systems with Discrete and Delayed Communication

Wan, W., University of California, Davis

The fundamental and practical challenges associated with the control of networked process systems have been the focus of significant research activities in process control over the past two decades and have motivated many research studies in this area. Examples include results on traditional centralized and decentralized control, distributed model predictive control, passivity-based control, agent-based systems, and singular perturbation formulations. In addition to these works, research has recently begun to address communication issues (e.g., resource constraints, real-time scheduling constraints, etc.) in the plant-wide control problem, leading to the design of resource-aware networked plant-wide control systems with explicitly-characterized stability and performance properties [1,2]. These efforts have been motivated by the increased reliance in the process industries on sensor and control systems that are accessed over shared communication networks instead of dedicated links, as well as the growing calls for expanding the traditional process control and operations paradigm in the direction of smart plant operations.

An important consequence of the tight integration between the different process subsystems (through inventory recycle) is the need for a similarly tight integration of the information flow between the local control systems. A balance must be maintained, however, between the desired control quality and the extent of information transfer. On the one hand, the need to account for the interactions between the different units and minimize the propagation of process upsets demands increased levels of local data collection and broad dissemination throughout the plant. On the other hand, the costs associated with sensor deployment and communication, together with the difficulty of obtaining frequent or timely measurements for certain process variables, impose constraints on the sensing and communication.

An approach that reconciles these conflicting objectives is model-based quasi-decentralized control, which is a distributed control strategy in which the local control systems rely on predictive models of the plant units to compute the local control action at times when communication is suspended, and communicate only at discrete times (to reduce communication costs) to provide updates to the models' states. In this approach, it was assumed that the communication delays were negligible and that the data transmitted by each subsystem was immediately available to the target control systems. In distributed control architectures, however, various forms of delays in measurement transmission are inevitable. These include communication delays (e.g., due to varying battery levels in a wireless sensor network) and processing delays (e.g., due to the use of computationally sophisticated data processing techniques). Such delays may degrade the closed-loop performance and even cause instability if not taken into consideration in the plant-wide control system design.

This paper focuses on networked control of multi-unit distributed plants subject to discrete and delayed communication between the local control systems. A model-based quasi-decentralized networked control structure that enforces closed-loop stability with minimal data transfer, while simultaneously taking into account the different communication delays associated with the transfer of information between the constituent subsystems, is developed. In this structure, each control system includes: (1) a set of predictive models each of which predicts the evolution of the state variables of a given neighboring unit when measurements are unavailable and is updated when communication is restored, and (2) a set of propagation models each of which uses the delayed measurements from a given unit, together with the past values of the estimated local control action, to generate an estimate of the current state which is then used to update the state of the corresponding model. A combined discrete-continuous system formulation of the networked closed-loop plant is developed and analyzed, leading to an explicit characterization of the closed-loop stability properties in terms of the update period, the communication delay sizes, the degree of plant-model mismatch, and the selection of controller design parameters. Finally, the theoretical results are illustrated using a reactor-separator plant example.


[1] Sun, Y. and N. H. El-Farra, ``Resource-Aware Quasi-Decentralized Control of Networked Process Systems over Wireless Sensor Networks," Chemical Engineering Science, 69, 93-106, 2012.

[2] Sun, Y. and N. H. El-Farra, ``Quasi-decentralized Model-Based Networked Control of Process Systems,'' Computers and Chemical Engineering, 32, 2016-2029, 2008.

See more of this Session: Control and Estimation of Large Scale Systems
See more of this Group/Topical: Computing and Systems Technology Division