FoSSA Optimization-Based SVM Classifier for the Recognition of Partial Discharge Patterns in HV Cables

Area of Science:

• Electrical Engineering
• Artificial Intelligence
• Machine Learning

Background:

• Cable partial discharge (PD) pattern recognition is crucial for electrical equipment diagnostics.
• Traditional Support Vector Machine (SVM) classifiers face challenges in accuracy and generalization.
• Existing optimization algorithms for SVM parameters can be prone to local extremums and slow convergence.

Purpose of the Study:

• To enhance the classification accuracy and generalization performance of SVM for PD pattern recognition.
• To develop a hybrid optimization algorithm for tuning SVM kernel and penalty parameters.
• To improve the efficiency and robustness of the optimization process.

Main Methods:

• A Firefly Optimized Sparrow Search Algorithm (FoSSA) was developed by integrating firefly algorithm's local search capabilities into the Sparrow Search Algorithm (SSA).
• Circle-Gauss hybrid mapping was used for improved population initialization in SSA.
• A dynamic step-size strategy was incorporated into SSA to refine convergence accuracy.

Main Results:

• The FoSSA demonstrated superior optimization performance on six benchmark functions compared to standard algorithms.
• The proposed FoSSA effectively optimized SVM kernel function parameters and penalty factors for PD pattern recognition.
• The SVM classifier optimized by FoSSA achieved a high recognition accuracy of 97.5% for PD patterns.

Conclusions:

• The FoSSA significantly enhances SVM performance in cable PD pattern recognition.
• The hybrid approach overcomes limitations of standard SSA, improving local search ability and convergence.
• This method offers a promising solution for accurate and reliable electrical equipment fault diagnosis.