Fault detection and classification using Artificial Neural Network (ANN)
Various reasons contribute to the occurrence of transmission line faults, which lead to system outages. To prevent power supply failure, a transmission line failure must be detected and isolated as soon as possible. This study aims to establish a system for identifying and classifying the various ki...
Saved in:
| 主要作者: | |
|---|---|
| 格式: | Thesis |
| 語言: | English |
| 出版: |
2022
|
| 主題: | |
| 在線閱讀: | http://eprints.utm.my/id/eprint/99470/1/IrdinaNadhrahMohdMSKE2022.pdf |
| 標簽: |
添加標簽
沒有標簽, 成為第一個標記此記錄!
|
| 總結: | Various reasons contribute to the occurrence of transmission line faults, which lead to system outages. To prevent power supply failure, a transmission line failure must be detected and isolated as soon as possible. This study aims to establish a system for identifying and classifying the various kinds of faults that might occur in power transmission lines, by using the Artificial Neural Network method. ANN is one of the most effective methods for identifying and diagnosing faults that have occurred. To develop ANN, backpropagation was used as a learning algorithm that could handle large volumes of data. These were the detection and classification training methods used by Levenberg-Marquardt (trainlm) and Scaled-Conjugate Gradient (trainscg). In order to implement this project, the ANN, feed-forward-networks-with-backpropagation algorithms for each phase (voltage and current) are selected. Based on the simulation results, fault voltages and currents are measured and allocated as input data for ANN. The training data set consists of with faults and without faults on three-phase lines and ground. MATLAB Simulink was used in this study to run tests on 14-bus, 30-bus, and 57-bus systems to generate fault parameters. This project examines neural network output for Line-to-Line-faults, Single-Line-to-Ground faults, and Double Line-to-Ground faults. Then, import the data into MATLAB as input for ANN to-detect-and-classify-the fault. Three critical phases are used in ANNs: training, validation, and testing. The mean squared error (MSE), correlation coefficient, and confusion matrix are all used to evaluate the detection and identification network efficiency. The detection in the 14-bus system achieves a satisfactory MSE of 1.8857e-15, correlation of 1, and accuracy of 100%, indicating that the performance of the system is excellent. Meanwhile, the classification achieved a reasonable MSE of 0.48107, a correlation of 0.75643, and an accuracy of 80%, indicating that the system is acceptable. |
|---|