Optimasi Likelihood Function Suara Corona Discharge Menggunakan Model Normal Hidden Markov Sebagai Langkah Awal Deteksi Dini Kegagalan Isolasi

Abstract

Insulation failure due to high voltage phenomena such as corona discharge (CD) still occurs in many electrical systems in Indonesia. This is due to not being able to do early detection of CD. One form of CD is sound. As the first step for early detection of insulation failure in the form of clustering, a study is needed (in a 20 kV cubicle) that can optimize the sound characteristics of CD, which is the aim of this research. Based on observations on the needle-rod electrode 3 cm apart, the smallest breakdown was obtained at 34.3 kV. So that the classification of CD sound is set into 3 clusters starting from the cubicle voltage of 20 kV until before the breakdown occurs, namely 33 kV. The temperature in the cubical is between 27.5℃ - 35.3℃ and humidity ranges from 70% - 95%. Feature extraction was carried out using the linear predictive coding (LPC) method, then optimization of the likelihood function was carried out using the normal hidden Markov model which is expected to be used as a first step for early detection of insulation failure.