On the use of recurrent neuro-fuzzy networks for predictive control
Alireza Sadeghian, J.D. Lavers · 2002
This paper presents the application of recurrent neuro-fuzzy networks for the predictive control of nonlinear, multivariable, complex systems such as electric arc furnaces. The main objectives are to investigate the capability of adaptive neuro-fuzzy networks to predict the V-I characteristics of electric arc furnaces and to compare the performance of the proposed predictors with that of the feedforward neuro-fuzzy predictors. The novelties of this work are to propose the notion of approximate prediction and to implement it using a recurrent neuro-fuzzy structure suitable for long-term prediction. Successful implementations of recurrent neuro-fuzzy predictors are described and their performances are illustrated.