Klasifikasi Partial Discharge Pada Isolasi Stator Generator Tegangan Tinggi Menggunakan Artificial Neural Network

Afandi, Ahmad Mustofa (2025) Klasifikasi Partial Discharge Pada Isolasi Stator Generator Tegangan Tinggi Menggunakan Artificial Neural Network. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5022211006-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (5MB) | Request a copy

Abstract

Kondisi sistem isolasi stator berperan penting pada terjadinya down time generator tegangan tinggi. Generator tegangan tinggi yang bekerja mengalami tekanan elektris, tekanan mekanis, tekanan termal, serta adanya kontaminan yang dapat menyebabkan partial discharge pada isolasi stator. Terjadinya partial discharge akan memicu kecacatan yang lebih parah hingga terjadinya kerusakan total pada isolasi stator. PD pada isolasi stator generator diklasifikasikan menjadi end-winding surface discharge, end-winding corona discharge, slot discharge, delamination discharge, dan void discharge. Studi ini mengklasifikasikan kelima jenis PD isolasi stator generator menggunakan artificial neural network. Pada studi ini, PD isolasi stator generator dihasilkan dengan cara membuat kecacatan yang umum terjadi di sistem isolasi stator. Pengujian PD dilakukan menggunakan tegangan AC pada level tegangan 1,5 PDIV. Data pengujian PD diakuisisi menggunakan osiloskop digital, lalu diolah menggunakan Matlab untuk mendapatkan plot PRPD. Data PD direduksi menggunakan metode ekstraksi fitur statistik, fitur fraktal, dan fitur PCA. Fitur data PD yang telah didapatkan digunakan sebagai input pelatihan dan pengujian ANN dengan algoritma feedfoorward backpropagation memanfaatkan fitur fungsi patternnet di Matlab. Arsitektur jaringan ANN terdiri dari 1 input layer, 1 hidden layer, dan 1 output layer. Jumlah neuron di hidden layer divariasikan dari 6 hingga 10 neuron. Hasil plot PRPD menunjukkan bahwa setiap jenis PD isolasi stator generator tegangan tinggi memiliki karakteristik yang khas ditinjau dari pola bentuknya, magnitudo PD, persebaran fasa PD, dan jumlah titik PD. Hasil klasifikasi PD isolasi stator generator tegangan tinggi oleh ANN menunjukkan bahwa rata-rata akurasi tertinggi diperoleh menggunakan input fitur statistik sebesar 99,944%, diikuti fitur PCA sebesar 99,904%, dan fitur fraktal sebesar 97,696%. Hasil penelitian ini dapat berkontribusi sebagai acuan dalam pengembangan ANN untuk klasifikasi PD isolasi stator generator dan identifikasi awal penyebab kerusakan stator generator.
========================================================================================================================================
The condition of the stator insulation system plays an important role in the occurrence of high-voltage generator down time. Working high-voltage generators experience electrical stress, mechanical stress, thermal stress, and the presence of contaminants that can cause partial discharge in stator insulation. The occurrence of partial discharge will trigger more severe defects to total damage to the stator insulation. PD on generator stator insulation is classified into end-winding surface discharge, end-winding corona discharge, slot discharge, delamination discharge, and void discharge. This study classifies the five types of generator stator insulation PD using artificial neural networks. In this study, generator stator insulation PDs are generated by creating common defects in the stator insulation system. PD testing was performed using AC voltage at a voltage level of 1.5 PDIV. The PD test data was acquired using a digital oscilloscope, then processed using Matlab to obtain the PRPD plot. PD data is reduced using statistical feature, fractal feature, and PCA feature extraction methods. The PD data features that have been obtained are used as input for ANN training and testing with the feedfoorward backpropagation algorithm utilizing the patternnet function feature in Matlab. The ANN architecture consists of 1 input layer, 1 hidden layer, and 1 output layer. The number of neurons in the hidden layer was varied from 6 to 10 neurons. The PRPD plot results show that each type of high-voltage generator stator insulation PD has distinctive characteristics in terms of its shape pattern, PD magnitude, PD phase distribution, and number of PD. The classification results of high-voltage generator stator insulation PDs by ANN show that the highest average accuracy is obtained using statistical feature input of 99.944%, followed by PCA features of 99.904%, and fractal features of 97.696%. The results of this study can contribute as a reference in the development of ANN for the classification of generator stator insulation PD and early identification of the cause of generator stator damage.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Artificial Neural Network, Generator Tegangan Tinggi, Klasifikasi Partial Discharge, Stator, Artificial Neural Network, High Voltage Generator, Partial Discharge Classification, Stator
Subjects:	T Technology > TK Electrical engineering. Electronics Nuclear engineering T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK3401 Insulation and insulating materials T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK7 Turbogenerators. Steam-turbines T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK7882.P3 Pattern recognition systems
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Electrical Engineering > 20201-(S1) Undergraduate Thesis
Depositing User:	Ahmad Mustofa Afandi
Date Deposited:	27 Jul 2025 14:31
Last Modified:	27 Jul 2025 14:31
URI:	http://repository.its.ac.id/id/eprint/121855

Actions (login required)

View Item