Optimizing the Design of a LCC-S resonant converter for Wireless Power Transfer using an Artificial Neural Network

LCC-S resonant converter is a very promising topology to realize Wireless Power Transfer. The higher number of reactive components, compared with more classic topologies of wireless converter (such as the Series-Series compensated converter) gives a quite high degree of freedom in the design of the components themselves to achieve specific performances. To optimize the choice of the circuital parameters, an Artificial Neural Network (ANN) has been adopted. An equivalent simple circuit has been considered as tool to obtain quickly a large dataset to properly train the neural network. To validate the procedure, the performances (in terms of output voltage and efficiency) of the ANN-designed converter predicted by the simplified model, have been compared with Simulink results. It results that in nominal conditions (resonant frequency 150 kHz), the converter works as boost converter, granting 250 V of rms output voltage with a rms input voltage of 180 V with almost 95% of efficiency (verified also in Simulink environment). Furthermore, the performances of the ANNconverter are perfectly aligned with Simulink results in a frequency range wider than 40% of the nominal working frequency.