Frequency-Domain Maximum-Likelihood Estimation of High-Voltage Pulse Transformer Model Parameters

Research output: Contribution to journalArticle


This paper presents an offline frequency-domain nonlinear and stochastic identification method for equivalent model parameter estimation of high-voltage pulse transformers. Such kinds of transformers are widely used in the pulsed-power domain, and the difficulty in deriving pulsed-power converter optimal control strategies is directly linked to the accuracy of the equivalent circuit parameters. These components require models which take into account electric fields energies represented by stray capacitance in the equivalent circuit. These capacitive elements must be accurately identified, since they greatly influence the general converter performances. A nonlinear frequency-based identification method, based on maximum-likelihood estimation, is presented, and a sensitivity analysis of the best experimental test to be considered is carried out. The procedure takes into account magnetic saturation and skin effects occurring in the windings during the frequency tests. The presented method is validated by experimental identification of a 2-MW-100-kV pulse transformer.


  • Davide Aguglia
  • Philippe Viarouge
  • Carlos Martins
Research areas and keywords

Subject classification (UKÄ) – MANDATORY

  • Natural Sciences
  • Physical Sciences


  • High-voltage (HV) techniques, identification, maximum-likelihood (ML), optimization methods, pulse transformers
Original languageEnglish
Pages (from-to)2552-2561
JournalIEEE Transactions on Industry Applications
Issue number6
Publication statusPublished - 2013
Publication categoryResearch