Abstract
Gas-insulated systems (GISs) are known for their advantages over conventional air-insulated systems (AISs) for the purpose of power substations. However, GISs reliability are threatened by the presence of metallic particles inside GIS compartments. Once present, these particles normally cause partial discharges (PDs), whose worse effect is the complete breakdown of the system. It is essential to know the characteristics of such particles, especially related to their lengths and their positions in GIS compartment, because particle with specific length and position can be seen as a unique defect. This paper discusses the estimation of length and position of metallic particle inside GIS by utilizing PD patterns generated by particles of different lengths and positions. Experiments to acquire PD signals were performed using the well-established IEC 60270 method. Simple and powerful C4.5 algorithm was used to estimate length and position of metallic particle by recognizing various PD patterns collected from the experiments. It is found that C4.5 algorithm used in this study can achieve an average accuracy of 91.7% for this purpose, which is fairly high. Therefore, this study can be a contribution for the improvement of GIS reliability.