Power Transformer Fault Prediction based on Support Vector Machine


Téléchargements par mois depuis la dernière année

Plus de statistiques...

Hesabi, H., Nourelfath, M., Hajji, A. et Jung, T. (2023) Power Transformer Fault Prediction based on Support Vector Machine. Dans CIGI Qualita MOSIM 2023, Trois-Rivières, Québec, Canada DOI 10.60662/qx91-3041.

[thumbnail of CIGI_QUALITA_MOSIM_2023_paper_7057.pdf]

Télécharger (220kB) | Prévisualisation


Power transformers (PTs) play a crucial role in power generation, transmission, and distribution systems. Monitoring the health of these transformers is essential to ensure an uninterrupted power supply. Dissolved Gas Analysis (DGA) is a widely used technique to examine the condition of PTs. However, predicting dissolved gas content in PTs is challenging due to non-linearity, high dimensionality, and limited training datasets. This paper presents a novel approach to predict PT faults using the Support Vector Machine (SVM) algorithm based on DGA data. The proposed method employs SVM to achieve accurate and timely fault diagnosis, which is essential for preventing faults, as manual diagnosis is time-consuming and expensive. The real data does not include all the desired labels, so Gaussian simulation generates new data that provides all labels. The new data is generated using the Inverse Cumulative Distribution Function (ICDF) to convert the Gaussian samples to samples from the specified distributions. The proposed approach achieves a probabilistic output for the fault diagnosis of oil-immersed transformers, overcoming the limitations of traditional DGA methods that often provide inaccurate diagnosis results and cannot summarize the fault development rule inductively. The case study results demonstrate the effectiveness of the proposed approach in predicting PT faults. Furthermore, this paper contributes a new method that utilizes SVM based on DGA data, which can help maintenance managers detect faults accurately and promptly, contributing to the PT fault diagnosis field.

Type de document: Document issu d'une conférence ou d'un atelier (NON SPÉCIFIÉ)
Mots-clés libres: Power transformer Dissolved gas analysis Fault diagnosis Support vector machine Machine learning
Date de dépôt: 17 août 2023 14:55
Dernière modification: 11 sept. 2023 20:24
URI: https://collection-numerique.uqtr.ca/id/eprint/2069

Actions (administrateurs uniquement)

Éditer la notice Éditer la notice