<![CDATA[Reliable fault identification is crucial for maintaining the stability and smooth operation of modern power systems. However, the advancement of data-driven fault detection models is sometimes hampered by the scarcity of real-world fault data. This study aims to assess the effectiveness of machine learning-based fault diagnostic methods in the context of limited data, using synthetic operational data derived from power system simulations. The study employs a simulation-based quantitative experimental methodology, leveraging synthetic data to simulate several failure scenarios on transmission lines. Experimental findings indicate that while certain classifiers can achieve high accuracy, there are significant differences in performance consistency when faced with complex fault scenarios. Margin-based learning models demonstrate superior performance stability compared to alternative methodologies, especially when faced with limited training data. These results demonstrate that selecting the right method is crucial for machine learning-based defect diagnosis to perform well when data comes from synthetic sources. This study contributes to making power systems more reliable by making fault detection more accurate and efficient and ensuring that maintenance is carried out on schedule to reduce system downtime]]>
Copyrights © 2025
