UHF signal model parameters identification and reconstruction for partial discharge in substation

This paper constructed a model of the ultra-high frequency (UHF) signals generated from partial discharge (PD) based on the autoregressive-moving average (ARMA) model, and proposed an algorithm for estimating the model order and parameters using higher order cumulants. To verify our algorithm, the UHF signals is simulated by double exponential oscillation decay function mixed with Gaussian white noise and fixed frequency signals. We use the parameters of ARMA model to generate the corresponding Fourier transformation aptitude, and use bispectrum estimation to construct the phase information, and eventually the time-domain signals are obtained. The results show that the algorithm is able to reconstruct PD signals obtained in noise environment.