Search for collections on Undip Repository

PENGGUNAAN MACHINE LEARNING DENGAN METODE LONG SHORT TERM MEMORIES SEBAGAI BASIS CLEANING MAINTENANCE DAN PERFORMA KONDENSOR TURBIN UAP PEMBANGKIT LISTRIK TENAGA GAS DAN UAP

SYACHRIR, Guntur and Utomo, M.S.K. Tony Suryo and Christwardana, Marcelinus (2025) PENGGUNAAN MACHINE LEARNING DENGAN METODE LONG SHORT TERM MEMORIES SEBAGAI BASIS CLEANING MAINTENANCE DAN PERFORMA KONDENSOR TURBIN UAP PEMBANGKIT LISTRIK TENAGA GAS DAN UAP. Masters thesis, UNIVERSITAS DIPONEGORO.

[thumbnail of Halaman Cover-1.pdf] Text
Halaman Cover-1.pdf

Download (221kB)
[thumbnail of Halaman Cover.pdf] Text
Halaman Cover.pdf
Restricted to Repository staff only

Download (599kB)
[thumbnail of BAB I.pdf] Text
BAB I.pdf

Download (342kB)
[thumbnail of BAB II.pdf] Text
BAB II.pdf

Download (577kB)
[thumbnail of BAB III.pdf] Text
BAB III.pdf
Restricted to Repository staff only

Download (272kB)
[thumbnail of BAB IV.pdf] Text
BAB IV.pdf
Restricted to Repository staff only

Download (702kB)
[thumbnail of BAB V.pdf] Text
BAB V.pdf
Restricted to Repository staff only

Download (213kB)
[thumbnail of DAFTAR PUSTAKA.pdf] Text
DAFTAR PUSTAKA.pdf
Restricted to Repository staff only

Download (952kB)

Abstract

Kondensor merupakan salah satu komponen penting dalam Pembangkit Listrik Tenaga Gas dan Uap (PLTGU) yang memengaruhi efisiensi sistem pembangkitan. Di PLTGU Priok, kegiatan pembersihan kondensor masih dilakukan berdasarkan jadwal waktu tetap (time-based), sehingga berisiko menyebabkan penurunan performa jika kondensor kotor sebelum jadwal pembersihan. Penelitian ini bertujuan untuk merancang sistem prediksi waktu pembersihan kondensor secara optimal menggunakan metode Long Short-Term Memory (LSTM), yaitu salah satu algoritma machine learning yang cocok untuk memproses data deret waktu. Data yang digunakan meliputi parameter operasi kondensor seperti suhu masuk dan keluar air laut, suhu uap buang, dan tekanan vakum, yang dikumpulkan dari sistem digital pembangkit selama dua tahun. Model LSTM dilatih untuk memprediksi nilai efisiensi kondensor dan mendeteksi saat performa menurun, dengan akurasi evaluasi sebesar R² = 0,89 dan MAPE = 7,72%. Hasil prediksi ini ditampilkan dalam dashboard berbasis PI Vision sebagai sistem peringatan dini (early warning system) bagi operator. Dengan sistem ini, pemeliharaan kondensor dapat dilakukan secara berbasis kondisi (condition-based), sehingga membantu meningkatkan efisiensi pembangkit dan mencegah kerugian akibat kerusakan atau derating yang tidak terdeteksi. Pendekatan ini juga sejalan dengan inisiatif digitalisasi pembangkit dari PLN Indonesia Power.
Kata kunci: PLTGU, kondensor, machine learning, LSTM, maintenance, efisiensi

The condenser is a crucial component in a Combined Cycle Power Plant (PLTGU) that significantly affects the overall efficiency of the power generation system. At PLTGU Priok, condenser cleaning is still performed on a fixed time-based schedule, which can lead to performance degradation if fouling occurs before the scheduled maintenance. This study aims to develop a predictive system for optimal condenser cleaning using the Long Short-Term Memory (LSTM) method, a machine learning algorithm well-suited for time-series data. The data used in this research includes operational parameters such as seawater inlet and outlet temperatures, exhaust steam temperature, and vacuum pressure, collected over a two-year period from the plant’s digital monitoring system. The LSTM model is trained to forecast condenser efficiency and detect performance decline, achieving an evaluation accuracy of R² = 0.89 and MAPE = 7.72%. The prediction results are integrated into a PI Vision-based dashboard, serving as an early warning system for plant operators. This approach enables condition-based maintenance for condensers, helping to improve power plant efficiency and prevent losses due to undetected failures or derating. It also supports the digitalization initiatives promoted by PLN Indonesia Power.
Keywords: combined cycle power plant, condenser, machine learning, LSTM, maintenance, efficiency

Item Type: Thesis (Masters)
Uncontrolled Keywords: PLTGU, kondensor, machine learning, LSTM, maintenance, efisiensi
Subjects: Sciences and Mathemathic
Divisions: Postgraduate Program > Master Program in Energy
Depositing User: ekana listianawati
Date Deposited: 09 Oct 2025 08:09
Last Modified: 09 Oct 2025 08:09
URI: https://eprints2.undip.ac.id/id/eprint/39854

Actions (login required)

View Item View Item