Accelerating frequency domain dielectric spectroscopy measurements on insulation of transformers through system identification

Abstract

Frequency domain dielectric spectroscopy (FDS) is a non-destructive method widely used to evaluate the state of dielectric materials. Traditional FDS measurement is conducted in frequency sweep mode between 1 kHz and 1 mHz. However, this method requires long measurement time when sweeping in low frequencies (1 mHz–1 Hz), thereby limiting its field applications. Hammerstein model is selected as the mathematical model for insulation systems according to known property of dielectric and established based on system identification theory. An equivalent mathematical model fitting >97% with the insulation system is obtained. The equivalent model can be used to calculate dielectric response parameters and construct an FDS. The proposed method requires 10 min to obtain dielectric response parameters under any frequency excitation in the low-frequency section. The measurement time decreases by up to 90% compared with that of traditional FDS. Comparison experiments are conducted between system identification method and traditional FDS measurement applied to different insulation systems, such as insulation pressboards, testing transformers, and field power transformers. Results show that frequency responses are consistent between the equivalent model and insulation system, ensuring the accuracy of FDS curve obtained by the proposed method.