Intelligent Fault Classification During Power Swings: S-Transform Feature Extraction and SVM-Based Decision Framework for Modern Transmission Networks
Purpose: To ensure an accurate identification and discrimination between actual faults and power swing conditions, particularly those faults that are caused during the process of power oscillations. Design / Methodology / Approach: The proposed approach combines advanced signal processing with intelligent machine learning. In this regard, an S-Transform is used to process the current signals to extract discriminative features that identify changes in the system caused by power swings and faults. Subsequently, these features are provided to an SVM classifier to ensure accurate identification and classification of faults caused during power swings. The proposed approach is validated on a three-machine, nine-bus test system using PSCAD software, under different symmetrical and unsymmetrical fault conditions and during power oscillations. Research Limitation: The performance of the suggested method using real-time field data, diverse power system configurations, various levels of noise, and renewable power penetration is not considered. Finding: The simulation results demonstrate that the suggested S-Transform-based SVM method can clearly distinguish between power swings and actual faults, including symmetrical three-phase faults that have characteristics similar to power swings. Practical Implication: The suggested method can be successfully applied to modern numerical and intelligent-type power system protection relays to assist in decision-making during power swings. Social Implication: By enhancing the reliability and selectivity of power system protection, the proposed method also helps improve the stability and sustainability of the power system. This reduces the chances of widespread power outages, which in turn contributes to economic development. Originality / Value: This paper proposes a new hybrid approach to fault detection in power systems based on S-Transform feature extraction and SVM-based intelligent classification.