Optimization of a neural architecture for the direct control of a Boost converter

Abstract

In research related to control of DC/DC converters, artifi cial intelligence techniques are a great improvement in the design and performance. However, some of these tools require the use of trial and error strategies in the design, making it diffi cult to obtain an optimal structure. In this pa-per, we propose a direct control based on artifi cial neural network, whose design has been optimized using bio-inspired searching strategies, with the idea of optimizing simultaneously two different but important aspects of the network: architecture and weights connections. The control was successfully applied to a boost type converter. The results obtained allow us to observe the dynamic performance of the scheme, in which the response time and variation in the output voltage can be concluded that the criteria used for the control loop design were appropriate.