Optimization and AI in Power Electronics

An increasing number of real word engineering problems can hardly be solved by physical-based approaches or numerical simulations, due to high calculation efforts or inaccurate material properties and physical boundary conditions.
By using data-driven techniques and artificial intelligence on measured or simulated data precise predictions of complex physical processes can be provided:

■  State of charge (SoC) prediction of lithium-ion battery cells and modules using long short-term memory (LSTM) and WaveNet approaches

■  Electric circuit metamodeling via fully connected neural networks

■  Defect detection on measured and simulated images, for instance of power electronic devices, lithography masks, and wafers

SoC prediction of lithium-ion battery cells using physical-based and AI based models