Adaptive Update Method of Digital Twin for DC-DC Converter Based on Mechanism-Data Drive

In order to solve the problem of poor precision in traditional modeling of dc-dc converter, an adaptive update method of digital twin (DT) based on mechanism-data drive is proposed in this article to build a high-fidelity DT model. First, a DT modeling method based on hybrid drive of mechanism-data driven model is proposed. The data-driven model based on Gate Recurrent Unit (GRU) is constructed to compensate the mechanism model in parallel. The virtual-real error can be reduced to 0.11%, which effectively improves the precision of DT modeling. Second, in order to solve the problem of model tracking accuracy reduction caused by component degradation, parameter identification of DT model based on Genetic Algorithm (GA) is proposed. The relative error of the identification results is close to 5%, which can realize the adaptive updating of the model. Finally, the fault prediction model of convolutional neural networks-long short-term memory (CNN-LSTM) based on high fidelity DT model is constructed to realize degradation trend prediction and fault warning of electrolytic capacitors.