A neural network based approach to the regulation of DC/DC buck converters

This paper proposes the neural network controller as a viable alternative to the PI controller used in DC/DC converters of the buck type for voltage regulation. The PI controller, although robust and simple, requires a priori knowledge of the system characteristics and once designed for a specific load, its parameters remain fixed. The neural controller, in the on line mode has the ability to learn from experience, thus eliminating the need for a priori knowledge of the system dynamics. The neural network can adapt to variations in the load, and still allow the system to track a specific reference without redesign. Performance comparisons made with the standard PI regulator clearly bring out the superior performance of the neural network regulator