Toward Embeddable Digital Twins: A Voxel-Based Approach for Power Converter Temperature Monitoring

This article presents a novel voxel-based approach employing the finite-difference time-domain (FDTD) method for real-time 3-D temperature mapping of power converters. With this approach, the transient thermal behavior of power converters is provided via a digital twin (DT), offering an advanced temperature monitoring solution for the power converter health management. Its essential feature is digitizing the power converter with its components from the physical design to thermal simulation. Additionally, this approach addresses the computational demands of the FDTD method by incorporating an adaptive mesh technique, significantly reducing resources required for thermal simulation of the power converter. To ensure monitoring in real-time, the DT is implemented on an embedded system. This article conducts the analysis of the suitability of digital signal processors (DSPs) versus field-programmable gate arrays (FPGAs), concluding that the FPGA capabilities are ideal due to the synergy of the thermal solver parallelization and the data management achieved by the system on chip (SoC) platform, in charge of data management. The capabilities and potential of the proposed approach are demonstrated through comparisons between experimental data and simulation results for both static and dynamic cases. Furthermore, the digital twinning of a three-phase LLC converter is detailed as a paradigm for employing the DT as a real-time temperature monitoring technique. Results show that the proposed method achieves precision with an error no higher than 5%, demonstrating the fidelity of the proposed method to reduplicate the transient thermal behavior of complex power converters.