Distributed model predictive control

Abstract

This special issue is devoted to Distributed Model Predictive Control (DMPC), which is an emerging topic for scientific research. There are many open issues and several DMPC methods that have been proposed for different problem setups. The goal of this issue is to provide a state of the art snapshot of the development of DMPC methods and applications. To that purpose, six papers dealing with methodology and interesting applications are included here.Many systems, such as complex manufacturing, distribution networks or process plants, are often composed by multiple networked subsystems, with many embedded sensors and actuators. They are characterized by complex dynamics and mutual influences such that local control decisions made by single units have long-range effects throughout the system. This results in a very large number of problems that must be tackled for the design of an overall control system achieving the operating requirements in an optimal manner. When considering a control problem for a large-scale networked system, using Model Predictive Control (MPC) in a centralized fashion may be considered impractical and unsuitable because of the computational burden, scalability issues and communication bandwidth limitations, all of which make on-line, real-time centralized control infeasible. It is also inflexible against changes of network structure and the limitation of information exchange between different agents who might be in control of local subsystems. In order to deal with these limitations, DMPC has been proposed for control of such large-scale systems by decomposing the overall system into small subsystems.The first paper, titled Distributed MPC for resource constrained control systems, by Helton Scherer, Eduardo Camponogara, Julio Normey-Rico, José L. Guzmán and José D. Alvarez [1], proposes a DMPC for dynamically and input decoupled subsystems, equipped with local constraints and coupled control input constraints. The DMPC methodology manages energy resources for a set of consumer subsystems. The objective of the controller is to optimally distribute the allowable energy to the subsystems. The proposed methodology yields a distributed solution that converges to the optimum that would be obtained by a centralized controller. The proposed algorithm is flexible, allowing the inclusion of additional agents without increasing complexity. This optimal performance is achieved by expressing the problem in terms of slack variables and the global coupling constraint as a set of local subsystem constraints; thereby, favouring the application of DMPC. The controller is tested in a hardware in the loop configuration composed by a real solar thermal plant and a set of simulated users. The obtained results show that the proposed technique provides promising results.Stefano Riverso and Giancarlo Ferrari-Trecate, in their article Plug-and-Play distributed model predictive control with coupling attenuation [2], consider the control of a large-scale system composed of physically coupled linear subsystems that can...

Optimal combined wake and active power control of large‐scale wind farm considering available power

“…The MPC is a widely used control method in wind farm optimization. The basic principle of the MPC can be seen in [30]. In this section, an optimal combined wake and active power control method based on MPC is generated to achieve maximization of total available power of wind farm and ensure power demand from TSO over a given prediction horizon.…”

Compositional synthesis via a convex parameterization of assume-guarantee contracts

“…The load-frequency control (LFC) problem in power networks [9] is addressed in this example. A power network is a collection of areas, where each area has its own generator and users, and can transfer its excess power to other areas through a set of links.…”

“…Since only the departure from the nominal state matters, the state vector for subsystem i is written as x i,t = [∆δ i,t , ∆f i,t ] T , with ∆ denoting the divergence from the nominal value. The system dynamics from [9] is discretized using the Euler method,…”

Nonlinear Model Predictive Control of Dynamic Positioning of Deep-Sea Ships With a Unified Model

“…The PID control algorithm is a typical process control algorithm, which is simple and at low cost, but it may not be able to achieve the required control accuracy due to the limit of available control torque. In this context, many advanced modern control algorithms were presented, such as adaptive control [8], optimal control [9], intelligent control [10], model predictive control (MPC) [11] and so on. Among them, the adaptive control, optimal control and intelligent control are almost based on the intelligent algorithms.…”