Adaptive Prony Method for Spectrum Estimation of Non-stationary Signals in Traction Systems

The spectral components of power static converters used for both locomotive drives and auxiliary services can be highly time-varying with consequent difficulties in their detection. In this paper an adaptive Prony method is proposed to calculate time-varying spectral components with good accuracy and acceptable computational efforts. The method is based on an adaptive technique which minimizes the mean square relative error of signal estimation. The application of the method to an auxiliary static converter shows its high accuracy and acceptable computational efforts