SISO System Model Reduction and Digital Controller Design using Nature Inspired Heuristic Optimisation Algorithms

Abstract

This research article explores an algorithm to reduce the order of a SISO system and thereby to design a digital controller. The reduced order modelling of a large complex system eases out the analysis of the system. AGTM (Approximate Generalised Time Moments) method was implemented wherein the responses were matched at different time instants to achieve the reduced system. This research work devises a new method, Ensemble Framework for Optimized System (EFOS), resulting into a reduced system with better performance as compared to conventional techniques. The research also efforts towards effective utilization of various heuristic algorithms like Genetic Algorithm, Particle Swarm Optimization and Luus Jaakola Algorithm, their implementation and a comparison with other techniques based on relative mean square error and time complexity. It was observed that the proposed transfer learning based approach, EFOS, combining the advantages of Luus Jaakola and Genetic algorithms depicted better results than their individual counterparts on diverse performance parameters like speed of convergence and optimal convergence to global minima. The percentage improvement achieved in the time taken for design of the digital controller was 85.3%, with no change in delta value.