City Research Online

Abstract

This thesis is concerned with the on-line determination of the optimum operating condition of some systems where the optimum solutions are computed using their simplified mathematical models whose parameters are periodically updated by comparing models and real processes outputs. In each case two problems have to be considered. Firstly, a parameter estimation problem to determine model parameters and, secondly, a system optimisation problem where the model is employed to determine the optimum operating condition to optimise a given performance criterion. In general, the mathematical model will not completely represent the structure of the real system and/or it is deliberately simplified in order to facilitate the computation of the optimal input condition and involve the estimation of fewer parameters, Hence, the two problems interact and several iterations between parameter estimation and system optimisation will be required before the final correct optimum condition is achieved. By treating the two problems of system optimisation and parameter estimation as two interacting subproblems within an integrated scheme and then applying hierarchical system theories of decomposition and subsystem coordination, several alternative techniques which take account of the interaction between the subproblems have been developed. Particular attention is given to a procedure which may be regarded as a modification of the two-step approach, commonly employed in industrial practice, in which the optimisation problem and the parameter estimation problem are treated separately and solved repeatedly until a final converged solution is obtained. The modified technique results in extra terms added to the optimisation objective function to ensure that the correct optimal operating condition is finally achieved on the real process in spite of model inaccuracies.

The technique is shown to perform satisfactorily in simulation studies of the following cases:

(i) Simulation of a chemical plant process where feed flow rate, reaction temperature and recycle ratio controller set points are adjusted to maximise the net rate of return, It has been assumed that the plant achieves its new steady state condition in between each change of set points, so enabling a steady state model to be used for the simulation investigation. In addition, the simulation is employed to investigate the effects of real process measurement errors.

(ii) Development of the above investigation for consideration of how the dynamics of the plant affect the performance of the algorithm.

(iii) Extension of the algorithm to an investigation of using a dynamic mathematical model in order to design a feedback control system where the gains are adjusted to optimise a given performance index.

(iv) Application of the design technique to the optimal feedback control of a synchronous generator system.

Publication Type:	Thesis (Doctoral)
Subjects:	Q Science Q Science > QA Mathematics
Departments:	School of Science & Technology > School of Science & Technology Doctoral Theses Doctoral Theses

Export

Downloads