TY - GEN
T1 - Prognostics Health Management (PHM) System for Power Transformer Using Kernel Extreme Learning Machine (K-ELM)
AU - Abdillah, Muhammad
AU - Krismanto, Awan Uji
AU - Nugroho, Teguh Aryo
AU - Setiadi, Herlambang
AU - Pertiwi, Nita Indriani
AU - Mahmoud, Karar
AU - Prasetio, Murman Dwi
N1 - Publisher Copyright:
© 2020 Association for Computing Machinery.
PY - 2020/9/28
Y1 - 2020/9/28
N2 - A power transformer is one of the most important and valuable components for the power system network. This device is critical to ensure power quality and reliable electricity supply for consumers. When the power transformer could not work properly or out of service in unforeseen ways, it provides a severe impact on power system utilities and customers in term of the expensive of transformer's replacement cost and revenue lost caused by the electrical blackout. To overcome these issues, the proper prognostics health management (PHM) system as a tool for condition monitoring and health assessment of these valuable assets is required. This paper proposed a PHM system based on a kernel extreme learning machine (K-ELM) for power transformer's health assessment. Two sets of variable combinations called Set-1 and Set-2 were considered to examine the robustness and efficacy of the proposed method. In Set-1, the input variables were water content, total acidity, breakdown voltage, dissipation factor, dissolved combustible gases, and 2-furfuraldehyde. While the output of PHM system was the health condition which categorized as good, moderate, and bad circumstances. Set-2 utilized water content, total acidity, breakdown voltage, dissipation factor, and interfacial tension as input variables. Whereas, the PHM system outputs consisted of four categories: normal, good, moderate, and bad. The proposed method with two sets of variables had showed the satisfactory results for transformer's health condition assessment compared to an extreme learning machine (ELM), support vector machine (SVM), and least-square support vector machine (LS-SVM) in terms of learning and testing accuracies and computation time. The proposed PHM system using the Set-1 dataset could assess the transformer health as of 100% while in terms of the testing process, the proposed PHM system has an excellent accuracy result as of 68.67%. Furthermore, the proposed PHM system using the Set-2 dataset had successfully assessed the transformer health as of 100%. In the testing phase, the proposed PHM system model has a rigorous result for its accuracy result as of 93.61%.
AB - A power transformer is one of the most important and valuable components for the power system network. This device is critical to ensure power quality and reliable electricity supply for consumers. When the power transformer could not work properly or out of service in unforeseen ways, it provides a severe impact on power system utilities and customers in term of the expensive of transformer's replacement cost and revenue lost caused by the electrical blackout. To overcome these issues, the proper prognostics health management (PHM) system as a tool for condition monitoring and health assessment of these valuable assets is required. This paper proposed a PHM system based on a kernel extreme learning machine (K-ELM) for power transformer's health assessment. Two sets of variable combinations called Set-1 and Set-2 were considered to examine the robustness and efficacy of the proposed method. In Set-1, the input variables were water content, total acidity, breakdown voltage, dissipation factor, dissolved combustible gases, and 2-furfuraldehyde. While the output of PHM system was the health condition which categorized as good, moderate, and bad circumstances. Set-2 utilized water content, total acidity, breakdown voltage, dissipation factor, and interfacial tension as input variables. Whereas, the PHM system outputs consisted of four categories: normal, good, moderate, and bad. The proposed method with two sets of variables had showed the satisfactory results for transformer's health condition assessment compared to an extreme learning machine (ELM), support vector machine (SVM), and least-square support vector machine (LS-SVM) in terms of learning and testing accuracies and computation time. The proposed PHM system using the Set-1 dataset could assess the transformer health as of 100% while in terms of the testing process, the proposed PHM system has an excellent accuracy result as of 68.67%. Furthermore, the proposed PHM system using the Set-2 dataset had successfully assessed the transformer health as of 100%. In the testing phase, the proposed PHM system model has a rigorous result for its accuracy result as of 93.61%.
KW - K-ELM
KW - LS-SVM
KW - PHMS
KW - Power System Network
KW - Power Transformer
KW - SVM
UR - http://www.scopus.com/inward/record.url?scp=85097331462&partnerID=8YFLogxK
U2 - 10.1145/3429789.3429822
DO - 10.1145/3429789.3429822
M3 - Conference contribution
AN - SCOPUS:85097331462
T3 - ACM International Conference Proceeding Series
BT - Proceedings of the International Conference on Engineering and Information Technology for Sustainable Industry, ICONETSI 2020
PB - Association for Computing Machinery
T2 - 2020 International Conference on Engineering and Information Technology for Sustainable Industry, ICONETSI 2020
Y2 - 28 September 2020 through 29 September 2020
ER -
