Feature Selection and Parameters Optimization of SVM Using Particle Swarm Optimization for Fault Classification in...

Area of Science:

• Electrical Engineering
• Computational Intelligence
• Power Systems

Background:

• Accurate fault classification is crucial for reliable power system operation.
• Existing methods may lack efficiency or accuracy in complex distribution networks.
• Radial distribution systems present unique challenges for fault detection.

Purpose of the Study:

• To develop a highly accurate and efficient fault classification technique for radial distribution systems.
• To leverage particle swarm optimization (PSO) for optimizing support vector machine (SVM) parameters.
• To enhance the selection of relevant input features for improved classification performance.

Main Methods:

• A particle swarm optimization (PSO) based support vector machine (SVM) classifier was developed.
• Time-domain reflectometry (TDR) with a pseudorandom binary sequence (PRBS) stimulus was used for dataset generation.
• The method was tested on a radial distribution network, identifying ten fault types using 12 input features from MATLAB and Simulink.

Main Results:

• The proposed PSO-based SVM classifier achieved a success rate exceeding 97%.
• The technique effectively selected optimal input features and tuned SVM parameters.
• High classification accuracy was demonstrated for various fault types in the test network.

Conclusions:

• The developed PSO-based SVM classifier is an effective and efficient method for electrical fault classification in radial distribution systems.
• This approach significantly improves classification accuracy, contributing to enhanced power system operational reliability.
• The integration of TDR and PSO-optimized SVM offers a robust solution for power system fault analysis.