On-line distributed hierarchical control and optimisation of large scale systems
This research is concerned with the application of closed-loop coordination techniques for on-line steady state optimisation and control of large scale systems using a micro-computer based system. A two-level hierarchical computer structure consisting of a coordinator at the supremal (upper) level and two local decision units at the infimal (lower) level had been established. Parallel computation were performed at the local decision unit level once the coordination parameters had been received from the supremal level. A steady state system model consisting of two interconnected subprocesses, simulated oy an analogue computer, was used to investigate the coordination methods for closed-loop hierarchical control. First-order time constants were introduced to the interaction inputs and the controls within the simulated subprocesses. Investigations had been carried out to study closed loop control using the Interaction Prediction and Interaction Balance coordination method. Special attention was given to the study of the problems associated with synchronisation of the local decision units for closed-loop control. Stability aspects of both coordination methods when subjected to disturbances in the controls and interconnections had been investigated~ Problems relating to system transient and local. decision asynchronisation, as well as their effects on system stability and convergence of the two tasks, namely the local decision task and the coordinator task had also been investigated. Methods for dealing with these problems had been suggested. The sub-optimality, convergency and robustness properties of each coordination method had been discussed. This research has demonstrated that the Interaction Prediction coordination methods are best suited for on-line distributed optimising control of large scale interconnected systems. Using the local feedback scheme, complete decentralisation at the local decision level operated asynchronou.sly can be achieved with the Interaction Prediction coordination method.