Abstract Modern electrical power networks have become increasingly complex, creating a strong need for fast and accurate identification
of system faults to ensure uninterrupted service (Kundur, 1994; Van Cutsem & Vournas, 1998). This research examines how
machine learning (ML) methods can improve the detection and classification of faults within power systems.
The study adopts a mixed research strategy, combining primary survey data with insights drawn from existing literature.
Various algorithms, including decision trees, support vector machines, and neural network models, are analyzed for their
effectiveness in detecting faults such as single line-to-ground, line-to-line, and three-phase faults (Samantaray, 2008;
Pradhan, 2010). Results indicate that ML approaches outperform conventional techniques by offering faster processing,
adaptability, and reduced dependence on manual operation (He & Yan, 2011).
Despite certain limitations such as high data requirements, computational complexity, and interpretability concerns, the
integration of ML shows strong potential in enhancing system stability and enabling advanced smart grid development
(Goodfellow et al., 2016).