Algorithms and tools for intelligent control of critical infrastructure systems

Abstract

Critical Infrastructure Systems (CIS) are spatially distributed and of a network structure. The dynamics are nonlinear, uncertain and with several time scales. There is a variety of different objectives to be reliably met under a wide range of operational conditions. The operational conditions are influenced by the disturbance inputs, operating ranges of the CIS, faults in the sensors and actuators and abnormalities occurring in functioning of the physical processes. The Chapter presents the intelligent multiagent structures and algorithms for controlling such systems. Each agent is an intelligent unit of high autonomy to perform the control functions over an allocated region of the CIS. Its mechanisms are structured in a form of a multilayer hierarchy. The regional agents are then integrated into the multiagent structure capable of meeting the operational objectives of the overall CIS. Several structures are considered starting from the completely decentralised with regard to the interactions between the local regions and ending up at the hierarchical architectures with the coordinating units, which are equipped with the instruments to coordinate activities of the agents across their functional layers. The required ability of the control system to meet the control objectives under a wide range of operating conditions is achieved by supervised reconfiguration of the agents. The recently proposed robustly feasible model predictive control technology with soft switching mechanisms between different control strategies is applied to implement the soft and robustly feasible agent reconfiguration. The generic ideas and solutions are illustrated by applications to two CIS: an integrated wastewater treatment plant and a drinking water distribution network.