ANALISIS SISTEM PROTEKSI DIFFERENTIAL RELAY TRANSFORMER (87T) PLTU TANJUNG JATI B MENGGUNAKAN METODE ARTIFICIAL NEURAL NETWORK (ANN)

Wahyu Micho Indrawan, Oskar Natan, S.ST., M.Tr.T., Ph.D.

2025 | Skripsi | ELEKTRONIKA DAN INSTRUMENTASI

Sistem kelistrikan harus menjaga kontinuitas pasokan listrik tanpa gangguan, termasuk mencegah trip tak terduga akibat gangguan internal maupun eksternal. Berdasarkan laporan gangguan No. 02/U3/RCFA/KPJB/2024, terjadi tiga kali trip pada differential relay transformer (87T) Unit #3 di PLTU Tanjung Jati B, yang menyebabkan blackout selama 20 jam. Sementara itu, Unit #4 dengan konfigurasi yang sama tidak mengalami gangguan. Hal ini menimbulkan dugaan adanya ketidaksesuaian antara pengaturan relay dengan kondisi aktual sistem. Penelitian ini bertujuan mengevaluasi kesesuaian setting sistem proteksi Unit #3 dengan pendekatan kecerdasan buatan.

Perancangan sistem analisis dilakukan melalui tiga tahap yaitu uji teoritis, simulasi konfigurasi relay, dan validasi data menggunakan metode Artificial Neural Network (ANN). Simulasi dilakukan pada MATLAB/Simulink dengan konfigurasi secara aktual. Model ANN dirancang dengan 13 parameter input yang mencakup tegangan, arus kerja tiap fasa, arus differential, dan arus restrain, serta satu output berupa status trip atau normal. Total 330 data latih dan 200 data uji digunakan untuk menguji performa klasifikasi gangguan oleh model ANN. Output yang didapat berupa korelasi antara hasil regresi dengan klasifikasi ANN.

Hasil menunjukkan ANN mampu mengklasifikasikan kondisi trip dengan akurasi 93,33?ri 330 data uji. ANN berhasil mendeteksi 3 kasus trip berulang sesuai kondisi aktual. Perbandingan parameter lainnya menunjukkan perbedaan signifikan antara setting aktual dan teori, seperti Slope 1 (25% vs 59%) dan Slope 2 (25% vs 119%). Simulasi juga menunjukkan waktu trip ±1725,8 ms masih dalam standar. Model ANN memiliki sensitivitas 99,3?n akurasi gangguan 87%, mengindikasikan bahwa penyebab utama trip lebih bersumber dari faktor eksternal. Penggunaan ANN terbukti meningkatkan keandalan dan selektivitas dalam kesesuaian setting sistem proteksi differential relay di PLTU Tanjung Jati B.

The electrical power system must maintain uninterrupted power supply continuity, including preventing unexpected trips caused by internal or external disturbances. According to disturbance report No. 02/U3/RCFA/KPJB/2024, three trip incidents occurred on the differential relay transformer (87T) of Unit #3 at PLTU Tanjung Jati B, resulting in a total blackout lasting 20 hours. Meanwhile, Unit #4, which has an identical configuration, operated without issues. This raises concerns about a possible mismatch between the relay settings and the actual system conditions. This study aims to evaluate the suitability of the protection system settings on Unit #3 using an artificial intelligence approach.

The system analysis design was carried out in three stages: theoretical testing, relay configuration simulation, and data validation using the Artificial Neural Network (ANN) method. Simulations were conducted in MATLAB/Simulink based on actual system configurations. The ANN model was designed with 13 input parameters, including voltage, phase current, differential current, and restrain current, with one output indicating trip or normal status. A total of 330 training data and 200 testing data were used to evaluate the ANN model's classification performance. The expected output includes the correlation between regression results and ANN classification.

The results show that the ANN model successfully classified trip conditions with 93.33?curacy across 330 testing data. It accurately identified 3 repeated trip events consistent with real conditions. Other parameter comparisons revealed significant discrepancies between actual and theoretical settings, such as Slope 1 (25% vs 59%) and Slope 2 (25% vs 119%). The simulation also showed a trip time of approximately ±1725.8 ms, which remains within operational standards. The ANN model demonstrated a high sensitivity of 99.3% and a fault classification accuracy of 87%, indicating that the primary cause of the trip is more likely external rather than internal. Thus, the use of ANN has proven effective in improving the reliability and selectivity of the differential relay protection system at PLTU Tanjung Jati B.

Kata Kunci : Differential Relay, ANN, MATLAB, Sistem Kelistrikan PLTU Tanjung Jati B