(214c) Robust Quasi-Decentralized Control of Networked Process Systems Under Communication Scheduling

Control of large-scale plants with distributed interconnected processing units is a fundamental problem in process control that has been the subject of significant research work in both the academic and industrial circles. Methodological frameworks that have been pursued to address this problem range from traditional centralized and decentralized control methods to distributed cooperative model predictive control formalisms to agent-based systems and supervisory control formulations. Recently, a quasi-decentralized networked control framework that offers a compromise between the complexity of centralized control designs and the performance limitations of fully decentralized controllers, has been developed in [1,2] for multi-unit plants whose subsystems communicate over a resource-constrained communication medium. In this framework, the transfer of information between the distributed control systems is kept to a minimum by allowing the local controllers to rely on the predictions of locally-embedded models of the other subsystems, and forcing communication to take place only at discrete time instances to provide the necessary correction to model prediction errors. This approach aims to provide a balance between the need to enforce the desired stability and performance properties on the one hand, and the need to conserve network resources, on the other. This is an important consideration, especially in view of the growing emphasis on augmenting process control systems with wireless sensor networks in order to expand their capabilities.

Beyond suspending and re-establishing communication, an important means of conserving network resources is to allow only a subset of the plant units to communicate with the rest of the plant at any given time according to a certain communication schedule. Forcing the different subsystems to transmit their data at different times (rather than simultaneously) creates an opportunity for providing a more targeted correction to the models' estimation errors, whereby the models with the largest uncertainties could receive more timely updates than would be feasible in the absence of scheduling. The successful implementation of this approach, however, requires that certain practical implementation issues, such as the presence of time-varying external disturbances, which arise frequently in process operations, be explicitly accounted for. For quasi-decentralized networked control systems, external disturbances, if not accounted for, may not only harm the stability and performance properties of the local control systems, but can also alter the optimality properties of the communication and scheduling policies, and ultimately interfere with the objective of balancing the inherent tradeoff between stability and network utilization.

The objective of this work is to develop a robust scheduled quasi-decentralized networked control framework for multi-unit plants that exchange information over a shared communication network and are subject to time-varying external disturbances. The framework integrates local model-based feedback control with sensor scheduling to robustly stabilize the plant while minimizing the rate at which each subsystem must collect and disseminate its data to the rest of the plant. The plant-wide control structure consists of a collection of robust feedback controllers that enforce an arbitrary degree of disturbance attenuation. Communication between the controllers is suspended periodically for time intervals during which each controller relies on models of the plant units to generate the necessary control action. Communication is then re-established at discrete time instances according to a certain schedule that determines the order and times at which the subsystems transmit the data needed to update the states of the models embedded in the target units. An explicit characterization of the interdependence between the minimum allowable communication rate, the scheduling strategy for sensor transmissions, the size of the disturbances, the accuracy of the models, the choice of control laws, and the achievable degree of disturbance attenuation, is obtained. Finally,, the results are illustrated using a reactor-separator plant example.

References:

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

[2] Sun, Y. and N. H. El-Farra, "A Quasi-decentralized Approach for Networked State Estimation and Control of Process Systems," Industrial and Engineering Chemistry Research, 49, 7957–7971, 2010.