Phy-caliper: Discovering Unknown Physics for Calibrating Predictive Maintenance in Power Electronics

Power electronic systems are the backbone energy router for renewable energy. Predictive maintenance is crucial to the safety operation of systems, which highly depends on the accuracy of modeling and simulation. The system model can be characterized by the well-established circuit theories in the power electronics field since the 1900s. However, when it comes to practical implementation, the complex but important effects of field external factors still cannot be explicitly characterized with existing knowledge. This project proposes an unorthodox and innovative idea to automatically discover and compensate for the complex field external factors when understanding the system is not sufficient, by using state-of-the-art deep learning concepts and tools. This data-driven knowledge discovery pipeline will lead to new theoretical insights in the field of power electronics. It would be a game changer to establish a new simulation standard for power electronics and beyond.