Transient Response Improvement of Repetitive-trained Neural Network Controlled DC-DC Converter with Overcompensation Suppression

This paper studies about a neural network control for digitally controlled method for PWM dc-dc converters. Especially, we focus on an overcompensation phenomenon which causes from a neural control term for the transient response improvement. It is known that a nonlinear prediction can improve the transient response, however, it also causes the overcompensation since the non-linear behavior mainly due the time delay of the converter system and digital control computation. In this study, an estimated current information from a steady state equation of the output voltage is used to suppress the overcompensation by the neural network control. The suppression method is simply realized by a mode change of the neural network control term in the transient state. From simulation results, it is confirmed that the proposed method can work both to improve the transient response and to suppress the overcompensation phenomena simultaneously.