Adaptive Digital Twin for DC-DC Boost Converter Using Recurrent Neural Network

DC-DC converters are extensively used in power electronics applications such as electric vehicles and renewable systems, where predictive modeling and real-time monitoring are essential for ensuring stability and performance. This paper describes the use of a Recurrent Neural Network (RNN) trained on MATLAB Simulink data to create an adaptive digital twin for a DC-DC boost converter. To facilitate real-time estimate and optimisation, the digital twin functions as a virtual model that precisely forecasts the internal states of the converter. The suggested method uses cascaded Simulink inputs and outputs to train the RNN, which enables the model to accurately represent dynamic system behaviour. By using a closed-loop boost converter and better training methods, error reduction is accomplished, and prediction accuracy is greatly increased in comparison to the open-loop scenario. In the closed-loop arrangement, the RMSE loss was seen to reduce, indicating the adaptive digital twin's reliability.