Digital Twin Based Identification of Passive Parameters of Three-phase Boost Rectifier using a GRU Neural Network

Condition and health monitoring in power electronics are becoming more and more important and are increasingly discussed in articles in the literature. To timely intervene in fault detection or remaining useful life (RUL) estimation for predictive maintenance operations, artificial intelligence (AI) and neural networks (NNs) can be useful also to estimate the health status of the components. In this paper a digital twin-based parameters identification method of an AC-DC converter is proposed using the Gated Recurrent Unit (GRU) NN. The dataset has been created without the need of any additional hardware, using only the sensors already present in the power conversion system to perform the control action. The training of the GRU NN has been performed in Matlab environment and the obtained results of the predicted parameters are presented to validate the proposed monitoring method. Furthermore, the closed form of the GRU NN has been implemented in the microprocessor present on the control board PED-Board through LabVIEW Real-Time graphical programming code. Finally, the experimental results of the parameters identification in real-time are presented.