PARTIAL DISCHARGE DEFECT PATTERN RECOGNITION ON POWER TRANSFORMER USING RANDOM FOREST
<p align="justify">Power transformer is a high voltage equipment that have a very important role in an electric power system. Failures on transformer can cause a significant problem in electrical power system. Partial discharge (PD) is one of electrical phenomena which might occur in...
Saved in:
| Main Author: | |
|---|---|
| Format: | Theses |
| Language: | Indonesia |
| Online Access: | https://digilib.itb.ac.id/gdl/view/28157 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | <p align="justify">Power transformer is a high voltage equipment that have a very important role in an electric power system. Failures on transformer can cause a significant problem in electrical power system. Partial discharge (PD) is one of electrical phenomena which might occur in power transformer. PD detection is an efficient diagnosis method to prevent the failure of electric equipment caused from degrading insulation. The appearance of partial discharges could indicate insulation aging or degradation and in long term may further reduce the integrity of the insulation, leading to the failure of the transformer. Random forest (RF) method is using to create a classifier model for PD defect pattern recognition on power transformer. PD measurements were conducted using conventional method as describe in IEC 60270. A commercial PD measurement system with detecting impedance is using to acquire PD signal from artificial defect. The output of this measurement is PRPD patterns. Totally there are 8 statistical PD features that extracted from PRPD pattern to identify the defect. Those features are PD amplitude, skewness, kurtosis, and variance which analyze on each half positive and negative cycle. Artificial defect use in this research are protrusion defect, floating metal defect, and void defect. To create a classifier model for PD defect pattern recognition using RF, a training data set and test data set was build using different data set from PD measurement. There are two classifier model creates in this research. The first model creates using all 8 PD features, while second model only use 4 PD features which selected by variable importance function of RF. The goal of reducing PD features using in classifier is to compare the accuracy of recognition with different features number. The result of this research shows that both classifier have 94.44% accuracy for defect pattern recognition. With this result it is shown that RF was successfully recognize different PD defect type on power transformer using 8 and 4 PD features with high accuracy. <p align="justify"> |
|---|