Analisis Temperatur Untuk Monitoring dan Diagnosis Kondisi Kesehatan Transformator Distribusi Berbasis Kecerdasan Komputasional

Rosmaliati, Rosmaliati (2022) Analisis Temperatur Untuk Monitoring dan Diagnosis Kondisi Kesehatan Transformator Distribusi Berbasis Kecerdasan Komputasional. Doctoral thesis, Institut Teknologi Sepuluh Nopember Surabaya.

Text 07111460010013-Dissertation.pdf - Accepted Version Restricted to Repository staff only until 1 November 2024. Download (5MB) | Request a copy

Abstract

Seiring meningkatnya beban daya listrik dan usia peralatan, memperpanjang masa manfaat transformator daya telah menjadi salah satu aspek terpenting untuk meningkatkan masa pakai infrastruktur sistem tenaga listrik, sambil mempertahankan keandalan sistem. Oleh karena itu pemantauan dan diagnosis transformator daya menjadi semakin penting, dimana pemantauan adalah pengumpulan data yang relevan selama layanan (on-line) atau selama periode pemeliharaan atau pengujian (off- line). Penelitian ini menerapkan beberapa metode pemantauan populer pada transformator menggunakan data Dissolved Gas analysis (DGA), temperatur dan parameter elektrik berbasis kecerdasan komputasional. Perbandingan beberapa metode untuk mendeteksi gangguan dengan data DGA menunjukkan bahwa Artificial Neural Network (ANN) lebih akurat dibandingkan dengan metode Decision Tree dan Random Forest. Pemodelan temperatur minyak transformator (top-oil) berdasarkan arus, pembebanan, dan faktor daya menggunakan Backpropagation Neural Network (BPNN) dan Radial Basis Function Neural Network (RBFNN) menghasilkan prediksi temperatur minyak transformator. Penelitian ini diterapkan pada transformator dengan kapasitas yang berbeda, dengan melakukan pelatihan dan pengujian data yang diverifikasi dengan data temperatur top-oil yang tersedia. Simulasi untuk memprediksi masa pakai transformator menggunakan Nguyen-Widrow Neural Network juga dilakukan dengan menggunakan data arus, temperatur, dan umur pada transformator. Data pelatihan dan pengujian adalah PSD (power spectral density), dan nilai energi yang dihasilkan dari proses wavelet, hasilnya menunjukkan bahwa metode algoritma Nguyen-Widrow dapat memprediksi masa pakai transformator lebih baik daripada BPNN. Pendekatan Extreme Learning Machine (ELM) untuk mengekstrak informasi maksimum dari transformator dapat memberikan perkiraan sisa umur transformator. Uji simulasi menunjukkan bahwa metode ELM menghasilkan MSE dan MAE terendah mengungguli metode prediksi lainya. Seluruh metode yang diterapkan menunjukkan bahwa metode kecerdasan komputasional dapat melakukan klasifikasi dan prediksi sisa umur yang berguna untuk proses monitoring dan diagnosis kondisi kesehatan transformator distribusi
======================================================================================================================================
As power loads increase and equipment life increases, extending the useful life of power transformers has become one of the most important aspects of increasing the service life of power system infrastructure, while maintaining system reliability. Therefore the monitoring and diagnosis of power transformers is becoming increasingly important, where monitoring is the collection of relevant data during service (on-line) or during periods of maintenance or testing (off-line). This study applies several popular monitoring methods to transformers using Dissolved Gas analysis (DGA) data, temperature and electrical parameters based on computational intelligence. Comparison of several methods for detecting disturbances with DGA data shows that the Artificial Neural Network (ANN) is more accurate than the Decision Tree and Random Forest methods. Modeling transformer oil temperature (top-oil) based on current, loading, and power factor using Backpropagation Neural Network (BPNN) and Radial Basis Function Neural Network (RBFNN) produces predictions of transformer oil temperature. This research was applied to transformers with different capacities, by conducting training and data testing which was verified with the available top-oil temperature data. Simulations to predict the lifetime of a transformer using the Nguyen-Widrow Neural Network are also carried out using current, temperature, and age data on the transformer. The training and testing data are PSD (power spectral density), and the energy values generated from the wavelet process, the results show that the Nguyen-Widrow algorithm method can predict the life of a transformer better than BPNN. The Extreme Learning Machine (ELM) approach to extract maximum information from the transformer can provide an estimate of the remaining life of the transformer. Simulation tests show that the ELM method produces the lowest MSE and MAE outperforming other prediction methods. All the methods applied show that the computational intelligence method can classify and predict the remaining life that is useful for monitoring and diagnosing distribution transformer health conditions

Item Type:	Thesis (Doctoral)
Additional Information:	RDE 621.314 Ros a-1 2022
Uncontrolled Keywords:	Transformator Distribusi, DGA, Neural Network, Extreme Learning Machine, Transformer Lifetime; Transformator Distribusi, DGA, Neural Networ, Extreme Learning Machine, Transformer Lifetime
Subjects:	T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK2551 Electric transformers.
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Electrical Engineering > 20001-(S3) PhD Thesis
Depositing User:	EKO BUDI RAHARJO
Date Deposited:	17 Apr 2023 06:54
Last Modified:	17 Apr 2023 06:54
URI:	http://repository.its.ac.id/id/eprint/97867

Actions (login required)

View Item