Implementation of a neural controller for the series resonant converter

A neural controller implementing an energy feedback control law is proposed as an alternative to classic control of resonant converters. The properties of the energy feedback control, and particularly the optimal trajectory control law, are analyzed. As a result, the state space is considered to be divided into two subspaces, that correspond to different states of the switches in the converter. An analog neural network learns to classify these two classes by means of a learning algorithm. A simple electronic implementation of this controller is proposed and applied to a series resonant converter (SRC). Results based on prototype measurements show a good improvement in the SRC response versus classical control methods based on the linearization of the state variable equations around a working point and confirm the validity of the neural approach.