DC-Link Voltage Wavelet Transform and Bispectral Analysis for Power Converter Voltage Unbalance Diagnostics

The power converter source voltage unbalance is the cause of unwanted voltage ripple that inflicts lower power quality and faster degradation of power electronics. As the voltage unbalance of the power supply progresses, the uncontrolled rectifiers switch the conduction mode from three to a single phase. This phenomenon can be detected using the information derived from the DC-link voltage signal. The paper presents two solutions for voltage imbalance detection and monitoring: the continuous wavelet transform and bispectral analysis. In comparison to the fast Fourier transform, which is one of the most common frequency domain analysis tools, both approaches provide clearer indicators with lower noise levels. The provided methods have undergone an initial simulation study with feasibility cheque and further experimental validation using a dedicated setup. Afterwards, a statistical analysis of 270 experimental datasets was conducted. The bispectral approach has more than 3 times higher signal-to-noise ratio than the Fourier transform, proving its noise robustness. As for the continuous wavelet transform, it has more than 2 times lower outlier percentage in comparison to classical frequency-domain analysis, making it a more stable algorithm.