Published Paper Details:

Power systems are prone to faults that can result in substantial damage to expensive components, explosions, outages, and even fatalities. To mitigate these consequences, a robust power protection system is crucial for detecting, classifying, and locating faults. This technical description presents a comprehensive methodology that utilizes machine learning algorithms such as Convolutional Neural Networks (CNN), Random Forest (RF), K-Nearest Neighbors (KNN), Decision Tree, and Support Vector Machine (SVM) to accurately detect and classify faults in electrical systems. The main objectives of this study are to identify and categorize various fault types occurring at different locations and resistance levels, gain insights into the causes of interruptions, restore power promptly, and minimize future occurrences. Additionally, the analysis aims to improve the understanding of protection system components to implement preventive measures and reduce the likelihood of service disruptions and equipment damage. The proposed solution integrates CNN as a machine learning algorithm to enhance fault detection and classification performance by leveraging its ability to extract relevant features from fault data.. The performance of various machine learning algorithms, including RF, KNN, Decision Tree, and SVM, is evaluated and compared. Among these algorithms, SVM demonstrates the highest performance in terms of fault detection and classification accuracy. It effectively identifies and categorizes different fault types, enabling swift fault location determination and subsequent mitigation actions. The experimental evaluation is conducted on a Test Network using MATLAB/SIMULINK, which provides a realistic representation of an electrical system. The results highlight the effectiveness of SVM in fault detection and classification, emphasizing its suitability for practical implementation in power system protection.