Review
Distributed control and optimization of process system networks: A review and perspective

https://doi.org/10.1016/j.cjche.2018.08.027Get rights and content

Abstract

Large-scale and complex process systems are essentially interconnected networks. The automated operation of such process networks requires the solution of control and optimization problems in a distributed manner. In this approach, the network is decomposed into several subsystems, each of which is under the supervision of a corresponding computing agent (controller, optimizer). The agents coordinate their control and optimization decisions based on information communication among them. In recent years, algorithms and methods for distributed control and optimization are undergoing rapid development. In this paper, we provide a comprehensive, up-to-date review with perspectives and discussions on possible future directions.

Introduction

Large-scale and complex systems have become prevalent in chemical and energy industries. Such systems arise from the integration of process units and plants as a result of sustainability motivations, specifically, economic efficiency, energy recycle, and carbon and water footprint reduction [1], [2]. Examples of such complex systems include chemical plants with recycles [3], [4], crude oil refineries [5], smart grids and microgrids [6], integrated heating, ventilation and air conditioning (HVAC) systems [7], and supply chains [8], [9].

Process control and optimization technologies are crucial to the implementation of such sustainable solutions in the chemical, energy and other relevant industry sectors. However, control and optimization become difficult for these large-scale and complex systems due to the size of their models and the existence of phenomena that emerge at a network level, such as multi-time-scale dynamics [10], multi-level uncertainties [11], and propagation of disturbances [12]. These challenges to the control and optimization of process networks call for theoretical efforts to deal with each individual phenomenon, and a computational architecture for control and optimization that accounts for their networked nature.

A promising approach to this end is to make the control and optimization decisions in a distributed manner. Specifically, a distributed architecture of control and optimization has the following meaning:

  • The entire process system, as a network, is decomposed into several constituent subsystems.

  • A computing agent (controller, optimizer) is associated with each subsystem and is responsible for making the decision in the corresponding subsystem.

  • Some information is exchanged between the agents to coordinate their decisions towards the control or optimization objective of the entire network.

A distributed architecture is different from a centralized one, in the sense that the decisions are made by multiple agents rather than a single agent. Compared to its centralized counterpart, distributed control and optimization are more scalable and parallelizable, and hence can be more efficient computationally. Moreover, a distributed architecture is flexible to the entering, exiting, and changes of subsystems (“plug-and-play” property [13]), and can be also combined with distributed fault detection for a fault-tolerant control architecture [14], [15]. It is also different from a decentralized architecture (which appears in traditional process control) due to the existence of some coordination mechanism, which generally enables the agents to account for other subsystems and arrive at desirable or optimal decisions for the overall network.

This paper provides a comprehensive and up-to-date review of distributed control and optimization. We aim to present the fundamental ideas and guide the readers through the most significant developments in recent years. The paper is organized as follows. We first introduce in Section 2 the preliminary concepts for the distributed architecture of control and optimization. Two major issues in developing a distributed control or optimization scheme, namely the methods of decomposing the network into subsystems, and the methods of performing optimization based on a network decomposition, are reviewed in Section 3 and Section 4, respectively. A case study on a reactor–separator process network is shown in Section 5. We discuss some important future directions in Section 6. Conclusions are made in Section 7.

Section snippets

Control and optimization

In the operations of process systems, many decisions need to be made. Decision making problems are organized in a hierarchy which contains the following problems on different time and space scales [8], [16]:

  • Control — input variables (flow rates, heat exchange rates, etc.) are manipulated to stabilize the system states (temperature, pressure, etc.) at their setpoints or track reference trajectories in a time scale of minutes to hours.

  • Real-time optimization — the setpoints or reference

Methods of Network Decomposition

The prerequisite to implement any distributed control or optimization scheme is to decompose networks into subsystems. That is, for control and optimization problems of process systems, the constituent objects in their definitions (inputs and states, optimization variables and/or constraints) need to be first assigned to different groups. Such a decomposition is not always trivial or arbitrary. As will be seen in the case study (Section 5), the network decomposition has a significant effect on

Methods of Decomposition-based Optimization

In the previous section we have presented methods of network decomposition. Once a decomposition is determined for the process, one can solve large-scale optimization problems by employing distributed optimization algorithms, which prescribe how the distributed agents update their decisions and how the information is transferred among the agents. In this section, we introduce the basics of the two major classes of formulations and algorithms of performing distributed optimization, namely block

Case Study on a Benchmark Process

In the works of our group, we have examined network decomposition and distributed optimization methods applied to the distributed MPC of a benchmark reactor–separator process [186], [187]. This system can be regarded as a minimal process network for distributed control and used for a demonstration of concept. For more complex processes, such as those appearing in oil refineries, in-depth investigations need to be carried out and reported.

As illustrated in Fig. 10, the process comprises two

Future Directions

In the previous sections, we have reviewed up-to-date methods of distributed control and optimization. For the future development and industrial application of distributed control and optimization methods, in our opinion, the following directions are of importance:

  • The communication network-induced issues need to be addressed in a more unifying framework that incorporates nonlinear systems and optimization-based, distributed agent decisions in a closed loop.

  • The network decomposition and

Conclusions

Large-scale, complex process networks are prevalent in chemical and other industries due to economics and sustainability considerations. Distributed control and optimization have been identified as a crucial decision-making paradigm for the operation of such complex networks, and studied extensively in the recent years. In this review paper, we first introduced the basic concepts of distributed control and optimization. We reviewed two most important aspects of distributed control and

References (200)

  • Y. Chu et al.

    Model-based integration of control and operations: Overview, challenges, advances, and opportunities

    Comput. Chem. Eng.

    (2015)
  • L.S. Dias et al.

    From process control to supply chain management: An overview of integrated decision making strategies

    Comput. Chem. Eng.

    (2017)
  • P. Daoutidis et al.

    Integrating operations and control: A perspective and roadmap for future research

    Comput. Chem. Eng.

    (2018)
  • I.E. Grossmann

    Advances in mathematical programming models for enterprise-wide optimization

    Comput. Chem. Eng.

    (2012)
  • I. Harjunkoski et al.

    Scope for industrial applications of production scheduling models and solution methods

    Comput. Chem. Eng.

    (2014)
  • P.M. Castro et al.

    Expanding scope and computational challenges in process scheduling

    Comput. Chem. Eng.

    (2018)
  • D.Q. Mayne

    Model predictive control: Recent developments and future promise

    Automatica

    (2014)
  • D. Hioe et al.

    Dissipativity analysis for networks of process systems

    Comput. Chem. Eng.

    (2013)
  • M.J. Tippett et al.

    Dissipativity based distributed control synthesis

    J. Process Control

    (2013)
  • R. Scattolini

    Architectures for distributed and hierarchical model predictive control — A review

    J. Process Control

    (2009)
  • P.D. Christofides et al.

    Distributed model predictive control: A tutorial review and future research directions

    Comput. Chem. Eng.

    (2013)
  • X. Ge et al.

    Distributed networked control systems: A brief overview

    Inf. Sci.

    (2017)
  • J. Liu et al.

    Distributed model predictive control of nonlinear systems subject to asynchronous and delayed measurements

    Automatica

    (2010)
  • H. Li et al.

    Distributed receding horizon control of large-scale nonlinear systems: Handling communication delays and disturbances

    Automatica

    (2014)
  • M. Heidarinejad et al.

    Handling communication disruptions in distributed model predictive control

    J. Process Control

    (2011)
  • I.G. Polushin et al.

    On the model-based approach to nonlinear networked control systems

    Automatica

    (2008)
  • E. Fridman et al.

    Control under quantization, saturation and delay: An LMI approach

    Automatica

    (2009)
  • Y. Ishido et al.

    Stability analysis of networked control systems subject to packet-dropouts and finite level quantization

    Syst. Control Lett.

    (2011)
  • F. Forni et al.

    Event-triggered transmission for linear control over communication channels

    Automatica

    (2014)
  • H. Li et al.

    Event-triggered robust model predictive control of continuous-time nonlinear systems

    Automatica

    (2014)
  • F.D. Brunner et al.

    Robust self-triggered MPC for constrained linear systems: A tube-based approach

    Automatica

    (2016)
  • P. Daoutidis et al.

    Decomposing complex plants for distributed control: Perspectives from network theory

    Comput. Chem. Eng.

    (2018)
  • K. Hangos et al.

    Optimal control structure selection for process systems

    Comput. Chem. Eng.

    (2001)
  • C. Ocampo-Martínez et al.

    Partitioning approach oriented to the decentralised predictive control of large-scale systems

    J. Process Control

    (2011)
  • P. Daoutidis et al.

    Structural evaluation of control configurations for multivariable nonlinear processes

    Chem. Eng. Sci.

    (1992)
  • S. Heo et al.

    Automated synthesis of control configurations for process networks based on structural coupling

    Chem. Eng. Sci.

    (2015)
  • L. Kang et al.

    Control configuration synthesis using agglomerative hierarchical clustering: A graph-theoretic approach

    J. Process Control

    (2016)
  • M. Moharir et al.

    Graph representation and decomposition of ODE/hyperbolic PDE systems

    Comput. Chem. Eng.

    (2017)
  • S.S. Jogwar et al.

    Networks with large solvent recycle: Dynamics, hierarchical control, and a biorefinery application

    AIChE J.

    (2012)
  • N.K. Shah et al.

    Short-term scheduling of a large-scale oil-refinery operations: Incorporating logistics details

    AIChE J.

    (2011)
  • M. Zachar et al.

    Microgrid/macrogrid energy exchange: A novel market structure and stochastic scheduling

    IEEE Trans. Smart Grid

    (2017)
  • M. Baldea et al.

    Dynamics and Nonlinear Control of Integrated Process Systems

    (2012)
  • D. Yue et al.

    Optimal supply chain design and operations under multi-scale uncertainties: Nested stochastic robust optimization modeling framework and solution algorithm

    AIChE J.

    (2016)
  • S. Riverso et al.

    Plug-and-play fault detection and control-reconfiguration for a class of nonlinear large-scale constrained systems

    IEEE Trans. Autom. Control

    (2016)
  • D.E. Seborg et al.

    Process Dynamics and Control

    (2010)
  • I.E. Grossmann

    Enterprise-wide optimization: A new frontier in process systems engineering

    AIChE J.

    (2005)
  • C.T. Maravelias

    General framework and modeling approach classification for chemical production scheduling

    AIChE J.

    (2012)
  • J.B. Rawlings et al.

    Model Predictive Control: Theory and Design

    (2009)
  • J.B. Rawlings et al.

    Fundamentals of economic model predictive control

  • M. Ellis et al.

    Economic Model Predictive Control

    (2016)
  • Cited by (14)

    • A stabilizing cooperative-distributed gradient-based economic model predictive control strategy for constrained linear systems

      2022, Journal of Process Control
      Citation Excerpt :

      The state-space (1)–(2) is denominated here as the original form of a state-space model description and can be obtained from a process plant through system identification techniques, as detailed by Gudi and Rawlings [26]. Additionally, if a centralized model is available, guidelines for its partition in subsystems are discussed in Mercangöz and Doyle [27], Al-Gherwi et al. [28], Rocha et al. [29], and Tang and Daoutidis [30]. During this partition, one can take into account the level of interaction among subsystems in order to reduce communication efforts.

    • Data-driven coordination of expensive black-boxes

      2022, Computer Aided Chemical Engineering
    View all citing articles on Scopus

    Supported by Division of Chemical, Bioengineering, Environmental and Transport Systems (CBET) of the National Science Foundation (NSF) of the United States of America.

    View full text