The identification of fuzzy grey prediction system by genetic algorithms

Abstract

The application of genetic algorithms to the identification of a fuzzy grey model is investigated. Based on a few past output data, the next output from the unknown plant can be predicted by the basic grey model. To improve the accuracy of the prediction model, a fuzzy controller is designed to determine the quantity of compensation for the output from the grey system. Genetic algorithms are used to optimize the roughly determined fuzzy model. A test pattern is then fed to the well-tuned fuzzy system to infer the quantity of compensation through the centre of gravity defuzzification method. The procedures of identifying three different types of fuzzy models are presented. Simulation results from a well-known example are used to demonstrate that simple modelling and accurate in prediction are the merits of the proposed methodology.