Evaluation of a neural control with optimal architecture for a DC/DC converter

The control of DC/DC converters, topologies widely used in the active reduction of harmonic distortion for nonlinear single-phase equipment of low power, presents great challenges of design, due to the complexity of the mathematical model and its highly nonlinear dynamic characteristic. Techniques of artificial intelligence such as neuronal networks, suppose great improvements in the design and the final performance, taking into account their abilities to learn complex dynamics and to generalize their behavior. The motivation to carry out this work is to propose a direct control loop with neuronal networks, (and then to evaluate the dynamic answer), allowing in addition to eliminate elements of test and error in the design. A direct control based on artificial neuronal network is proposed, whose design was performed in an excellent way using bio-inspired search patterns, in order to optimize simultaneously two different but important aspects of the network: the architecture and the weights of the connections. The control is applied to a boost converter. The obtained results allow to observe the dynamic performance of the scheme, in which the time response and the delta of output voltage allow to conclude that the criteria selected for the design of the control were appropriate, and represent a contribution in the development of control applications of DC/DC switched systems.