Differential dissipativity based distributed MPC for flexible operation of nonlinear plantwide systems

Abstract

Shifting away from the traditional mass production approach of the process industry, towards more agile, cost-effective and dynamic process operation, provides motivation for next-generation smart plants. The control system for smart plants needs to be capable of dynamically handling a wide range of operating conditions, whilst minimising operation costs during transitions, in addition to efficiently dealing with large-scale systems. This article presents a flexible and scalable Distributed Model Predictive Control (DMPC) approach based on differential dissipativity, which permits arbitrary cost functions (including economic costs). First, a plantwide contraction condition that ensures convergence to any feasible setpoint is constructed based on the process network topology and control synthesis of individual subsystems. This condition is then imposed as a constraint on local MPC controllers in a distributed manner, resulting in a scalable DMPC scheme where individual subsystems minimise arbitrary cost functions, whilst sharing the responsibility for plantwide stability.