Thermal digital twin

Thermal digital twin for power electronics

What is a thermal digital twin?

Thermal digital twins are simulation models running in parallel of a physical system, with its parameters dynamically updated through lifetime based on real-time measurements. This ensures optimal accuracy in replicating the behaviour of the physical counterpart.  

In the context of power electronics, thermal digital twins model the thermal behaviour of the power converter to estimate its temperature and overall state of health. By measuring a limited number of temperatures at selected points within a power module, a comprehensive equivalent thermal model is established. This model is then used to accurately estimate transistors’ temperatures and assess their level of degradation.  

Our technology offers you enhanced performance and minimized down-time for your hardware:
  • Online ageing monitoring: continuous observation of the thermal system degradation during online operation to avoid downtimes.
  • Optimised maintenance planning: utilize remaining useful lifetime estimation for efficient maintenance plan.
  • PQ-power curve extension:  extend the power quality curve by accurately estimating junction temperatures, enabling design margin reduction.
  • Reduced manufacturing costs and footprint: achieve cost savings and smaller footprint by minimizing design margins.
  • Optimised thermal management: implement dynamic control of the active cooling system for efficient thermal management.
  • Limited additional cost: provided by the limited extra sensing and computing power required.

© Fraunhofer ISIT
Excellent matching between measured & estimated temperatures

The evolution of distributed renewable energy sources, energy storage, as well as the rise of electromobility alongside with other mission-critical applications, presents significant challenges for power electronics systems. Power converters must reliably deliver electric energy with minimal loss, all within compact systems.

However, power electronics systems face inherent limitations in reliability and overcurrent capability, making them susceptible to overheating and thermal cycling stress. Achieving optimal performance requires effective thermal management strategies. Additionally, the implementation of condition monitoring with minimal sensing effort is crucial for efficient degradation monitoring and maintenance planning.

Offshore assets, such as wind turbines, are characterised by low operating margins  and high maintenance costs, particularly in the case of unplanned repairs. In this application, thermal digital twins can monitor the state of health of the inverter in real time to better plan the replacement of power modules and to identify the need to operate in degraded mode to avoid total failure.

Grid-connected energy storage systems must meet the highly dynamic, fluctuating power demands of the grid. In this application, thermal digital twins can help extract more power from the inverter. In fact, they are the optimal solution to provide an accurate estimation of the temperatures of the power devices, allowing the design margin to be minimised and the inverter to be pushed to the maximum of its capabilities. 

Industrial power supplies are expected to deliver power with high availability, under what may be a highly dynamic mission profile that can put stress on the power devices. Beside advanced powering solutions, our digital twin monitors operation and ensures that maintenance can be optimally planned to avoid unexpected failures, possible downtime and resulting operating loss.

For an online identification of thermal network, several algorithms such as particle swarm optimization, recursive least squares and dual extended Kalman filter (DEKF) were implemented and tested. The selection of identification algorithm is based on requirements, available computational resources, sensory, a power module topology and a use-case. The proposed digital twin model considers chip heating due to its power losses as well as due to the power losses of neighbouring chips via cross-coupling terms. Under quasi-steady state operating conditions, the identified model provides an exact match to measured waveforms.

Our team is pioneer in the field of thermal digital twin for power electronics, and our work has been published in high impact journals and presented in various conferences:

  • M. Votava, K. Debbadi, Y. Pascal, M. Liserre, “Multi-step Least Squares Algorithm for Thermal Characterization Based on Mission Profile”, Accepted for publication to the IEEE Applied Power Electronics Conference and Exposition, 2024. APEC'24.
  • M. Votava, Y. Pascal, M. Liserre and Z. Peroutka, "Utilization of Least Squares Algorithm for Online Identification of Foster Thermal Network Parameters," PCIM Europe 2023; International Exhibition and Conference for Power Electronics, Intelligent Motion, Renewable Energy and Energy Management, Nuremberg, Germany, 2023, doi: 10.30420/566091222.
  • J. Kuprat, K. Debbadi, J. Schaumburg, M. Liserre and M. Langwasser, "Thermal Digital Twin of Power Electronics Modules for Online Thermal Parameter Identification," in IEEE Journal of Emerging and Selected Topics in Power Electronics, doi: 10.1109/JESTPE.2023.3328219.
  • J. Kuprat, Y. Pascal and M. Liserre, "Real-Time Thermal Characterization of Power Semiconductors using a PSO-based Digital Twin Approach," 2022 24th European Conference on Power Electronics and Applications (EPE'22 ECCE Europe), Hanover, Germany, 2022