Applying Bayesian Optimization to Parameters Selection in Waveform Reconstruction of Time-Varying Components for Power Quality Study

The widespread power electronics-based renewable energy resources and modern nonlinear loads in power systems present challenges in accurately assessing power quality due to their produced time-varying voltage and current waveforms. While several time-frequency analysis methods have been proposed to overcome the limitations of traditional frequency domain analysis, parameter adjustments are often needed for different types of signals. This paper focuses on applying Bayesian optimization to select the suitable parameters for the short-time Fourier-based multi-synchrosqueezing transform. The results obtained from simulations and actual measurements provide an appropriate range of parameter settings that reduce assessed time and validate the performance of the algorithm. Moreover, two feature importance indices emphasize the relevance and significance of the parameters.