Multi-class Fault Analysis Using Quadratic SVM Classifier Fault Detection Technique for Micro-grid

Abstract

An innovative and intelligent method is used to identify and diagnose micro grid electrical faults in this paper. In a dynamic distribution network with integrated distribution generation (DG), or micro-grid, fault detection is crucial. Intelligent approaches can help the micro-grid discover issues independently. In our case, the microgrids employs batteries and solar panels in which faults in the grid integration of a solar and battery energy storage system (BESS) must be detected and established quickly. Due to the complexity of linking two-generation systems with the grid, which involves many parameters, the Simulink model simulates micro-grid normal operation and failure scenarios. The simulated fault circumstances depict distribution line network faults. To reach our objective of the study the Quadratic Support Vector classifiers will be compared to linear regression, Kernal Naives bayes, and sub-space discriminant intelligent detection methods. Training and testing these models will use feature-extracted voltage and current samples. The classification rate, expressed in statistical metrics, shows that quadratic support vector is more efficient and accurate than other classifiers which it tested and trained against, with the next best classifier 3.3% below it in accuracy. We substantiate our claims graphically with scatterplots and confusion matrices for each classifier.