Black-Box Large-Signal Average Modeling of DC-DC Converters Using NARX-ANNs

This paper investigates the use of non-linear autoregressive exogenous (NARX) artificial neural networks (ANNs) to achieve black-box average dynamic models dc-dc converters capable capturing main converter non-linearities. Non-linearities may include, for example, behavior variations due changes operating point or mode (e.g., discontinuous conduction mode, continuous mode). presents design guidelines determining NARX-ANN architecture and dataset be used in training process. Dataset definition includes choice perturbations stimulating aimed system behaviors optimizations size reduction. The proposed approach is first derived a boost converter. To verify generality method, same methodology also applied Ćuk In both cases, modeling provided accurate results, with only limited deviations observed time-domain responses step duty-cycle output current. model small-signal under different conditions. validity evaluated experimentally by considering prototype.