Neural-fuzzy control of DC-DC converters : An off-line learning approach

Conventional linear frequency domain control methods fail to provide nonlinear control to cope with possible large variable fluctuation in DC-DC converters. Owing to the nonlinear behaviour, mainly provided by the fuzzy rules and membership functions of fuzzy logic, fuzzy logic control is a good alternative for controlling DC-DC converters. The problem of designing fuzzy logic controllers for DC-DC converters involves determining the fuzzy rules, the associated antecedent membership functions and consequent membership functions. This paper discusses the methodology of using neural networks to adjust the shapes and locations of the membership functions so that an optimal fuzzy control can be achieved. This method can eliminate the need for heavy computation in multi-variable control. (author). 10 figs., 12 refs.