Implementation of Small-Scale Artificial Neural Networks for Power Electronics Applications

Inhalt:
Artificial Neural Networks (ANNs) can be used in power converter applications to identify a relationship between the input variables and target output variables dependent on these input variables. This feature can be used to integrate enhanced control capabilities into power converters. ANNs can be deployed in digital controllers running on power converters for real-time tasks. In such applications, the memory and computation requirements of the implemented ANN model become critical. In this study, different feed-forward ANN models are first implemented. Then, their inference durations and memory sizes are compared. Optimal phase-shift angle prediction using ANNs for PWM interleaving in multi-input single-output power converters is also addressed as a case study.