Partial Discharge Intensity Estimation on Hydrogen-Cooled Generator Using Fuzzy Logic Based Decision Tree Algorithm

• Published in: 2023 International Conference on Technology and Policy in Energy and Electric Power (ICT-PEP) pp. 18 - 23
• Main Authors: Arsya, Nanda Redha, Suwarno, Auluddin, M. Arief, Albab Fauzi, M Annas
• Format: Conference Proceeding
• Language: English
• Published: IEEE 02.10.2023
• Subjects: Couplings; decision tree algorithm; Dielectric materials; Fuzzy logic; generator; Hydrogen; partial discharge; Partial discharges; Prediction algorithms; Windings
• DOI: 10.1109/ICT-PEP60152.2023.10351147

Suralaya Power Generation Unit is one of the largest power generating units serving the 500 kV network in the Java-Madura-Bali (Jamali) area. The generator which is one of the main assets is the Hydrogen-cooled type generator. The healthiness of the generator can be determined from many parameters, including partial discharge. Partial discharge is a phenomenon of discharging electric charge partially in a dielectric material. Although partial discharge cannot diagnose whether the generator is damaged or not, this parameter is considered sufficient to determine the isolation conditions of the generator. The sensor that is commonly used to detect partial discharge in generators is a coupling capacitor, but the use of coupling capacitors is limited to the tip of the generator phase which leads to the insulated phase busduct (IPB). So that in some large-capacity generators, a resistance temperature detector (RTD) sensor is used by utilizing radio frequency to detect partial discharge throughout the generator windings. The results of this RTD reading can be represented in terms of Partial Discharge Intensity (millWatt), Pulse Count (Pulse per Second/PPS), and Qmax (mV). These quantities can be read in real-time to determine the condition of the dielectric material of the generator. In addition, based on practice in the field, the partial discharge intensity is often used in making decisions about whether the condition of the generator winding insulation is still good or not. This will be a problem if there are failures in reading the partial discharge intensity. Therefore, to overcome this problem, an analysis will be carried out using a fuzzy logic-based decision tree algorithm as fuzzy rules in estimating the value of the intensity of partial discharge using a reference Pulse Count, Qmax, and Pressure Hydrogen internal generator. Based on test results, this method can predict partial discharge intensity values quite accurately with an average deviation of 10.24 mW