Partial discharge classification on xlpe cable joints under different noise levels using artificial intelligence techniques / Wong Jee Keen Raymond
Cross linked polyethylene (XLPE) cables are widely used in power industries due to their good electrical and mechanical properties. Cable joints are the weakest point in the XLPE cables and most susceptible to insulation failures. Any cable joint insulation breakdown may cause a huge loss to power c...
| Main Author: | |
|---|---|
| Format: | Thesis |
| Published: |
2016
|
| Subjects: | |
| Online Access: | http://studentsrepo.um.edu.my/6719/ http://studentsrepo.um.edu.my/6719/4/jee_keen.pdf |
| Summary: | Cross linked polyethylene (XLPE) cables are widely used in power industries due to their good electrical and mechanical properties. Cable joints are the weakest point in the XLPE cables and most susceptible to insulation failures. Any cable joint insulation breakdown may cause a huge loss to power companies. Therefore, it is vital to diagnose the insulation quality to detect early signs of insulation failure. Partial discharge (PD) measurement is a vital tool for assessing the insulation quality at cable joints. Since the past, there have been many pattern recognition methods to classify PD, where each method has its own strengths and weaknesses. Although many works have been done on PD pattern recognition, it is usually performed in a noise-free environment. Also, works on PD pattern recognition are mostly done on lab fabricated insulators, where works using actual cable joints are less likely to be found in literature. Therefore, in this work, classification of real cable joint defect types using partial discharge measurement under noisy environment was performed. Five cross-linked polyethylene (XLPE) cable joints with artificially created defects were prepared based on the defects commonly encountered on site. A novel high noise tolerance principal component analysis (PCA)-based feature extraction was proposed and compared against conventional input features such as statistical features and fractal features. These input features were used to train the classifiers to classify each PD defect type. Classifications were performed using three different artificial intelligence classifiers, which include Artificial Neural Networks (ANN), Adaptive Neuro-Fuzzy Inference System (ANFIS) and Support Vector Machine (SVM). The performance of each classifier and feature extraction method was evaluated. It was found that SVM and ANN performed well while ANFIS classification accuracy was the weakest. As for input features, the proposed PCA features displayed highest noise tolerance with the least performance degradation compared to other input features. |
|---|