Stochastic Modelling of Power Electronic Converters under Uncertainties

The aim of this paper is to explore the usage of different sampling schemes in order to build a sparse Polynomial Chaos (PC) model for prediction of the oscillation frequency and its corresponding magnitude of the output voltage in a half-bridge buck converter. The 22 parasitic elements of this converter are modelled as random variables following a uniform distribution. As a result, from the considered sampling schemes, an Latin Hypercube Sampling (LHS) and Sobol’ sequences are determined to be the best in prediction of the magnitude, both with respect to accuracy and computational time needed for the creation of the model. The introduced accuracy measure, namely the a posteriori error, shows that the PC model approximates the oscillation frequency well. However, from a visual comparison with a model obtained from 8000 Monte Carlo samples, it can be seen that the PC model is not able to handle such a complex relationship properly.