Evolution computing and optimisation algorithm development: its applicant to system modelling and forecasting

Abstract

This project aims at unifying the mathematic process with evolutionary process and develop a new genetic operation so as to solving constrained optimization problems. The coevolutionary augmented Lagrangian method is a new idea so as to unify the mathematic process with evolutionary process. Base on the similar concept, a new genetic algorithm, which incorporate with the experimental design technique is developed so as to enhance the search of optimal solution and speed up the rate of convergence in genetic algorithm. The Taguchi genetic algorithm is executed to solve two sets of benchmark test problems. The results indicate that the rate of convergence is improved and quality of solution is enhanced. Moreover, this project proposes the new algorithm for mesh network link capacity assignment problem. As the link capacity assignment problem is a constrained optimization problem, results are evaluated and performance is compared with traditional genetic algorithm. It demonstrates the advantages of using Taguchi genetic algorithm in real world application.