This paper demonstrates improved fault detection rates in power systems using machine learning models, suggesting new adaptive strategies.
Key Points
The physics-informed neural network achieved a fault detection accuracy of 99.86%, showcasing its effectiveness in power systems.
A variety of models, including ANN, SVM, and LSTM, were developed to enhance classification sensitivity without fixed thresholds.
Robust performance was confirmed across various noise levels and training data percentages, indicating model reliability.
The intelligent system integrates advanced learning methods with traditional protection models, highlighting a notable improvement in operational efficiency.