Model Order Reduction of Large Scale System and Controller Design

Abstract

All systems available in the universe are expressed by their model, represented by their transfer function form or state-space form. In this developing era, the system s complexity is increasing, and hence the models are becoming complex. Due to the complexity, the transfer functions or state models lead to higher order models. So there is the requirement to design a suitable technique for reducing the order of the model of high order to reduce the complexity of the systems and reduce the analysis time of these systems. Numerous methods have been developed in recent years to aim that the error amongst the system of high order and reduced order is reduced to a lower value and maintaining all the characteristics of the system of high order in its approximation. The methods developed so far has several limitations due to which these methods are not applicable on all type of systems. Hence, a blended type of technique is developed in the thesis, which applies to all types of systems whether the model of the system is presented in the time domain or it exists in the frequency domain, whether it is continuous or discrete, whether it is SISO (Single Input Single Output) or MIMO (Multi Input Multi Output) model. This novel blended technique is formed by blending two techniques of reduction of the order of a model. These techniques are; an improved version of pole clustering and genetic algorithm. newline