Development of Adaptive Digital Twin for DC-DC converters using Artificial Neural Networks
Benjamin Jessie, Babak Fahimi, Poras T. Balsara
Abstract
This paper explores development of adaptive digital twins (DT) for DC-DC converters using artificial neural networks (ANN). The proposed DT can adaptively correct internal state prediction error for power converters modeled by a standalone recurrent neural network (RNN). To examine the proposed DT, DC-DC boost converters under open and closed loop modes of operation have been targeted. The neural network design and training process is explained, and the trained digital twin networks are successfully tested using exact time domain models of the converters. Finally, using a cascaded ANN the adaptive capability of the DT to account for RNN prediction errors due to varying systems inputs is successfully implemented.