Development of a new controller to optimize operation of a 150 watt, Half-Bridge DC-DC converter

This paper presents simulation, design methodology and implementation of a Half Bridge DC-DC Converter using an optimized multi-layer feed-forward Artificial Neural Network (ANN). The ANN controller provides an appropriate control signal to keep the output voltage of the converter stable. It was trained, using back-propagation algorithm according to an ideal MATLAB SIMULINK model of the converter. The Ideal model is run `off-line´ to generate a record of `perfect´ control signals in response to input and output perturbations. The recorded data then are used as an ´off-line´ training set for a Neural Network controller. A network pruning algorithm is used to reduce the size of controller and an online training process performed to fine-tune the controller parameters as well. Experimental results show that the ANN controller can guarantee an acceptable load and line regulation and provide improved performances. The proposed controller can be readily expanded to other topologies of DC converters.