Early Warning System Prediksi Remaining Useful Life Turbin Pada Demister Dengan Steam Quality NCG Menggunakan Metode Long Short Term Memory (LSTM)

Dewanti, Rosma Aprilia (2026) Early Warning System Prediksi Remaining Useful Life Turbin Pada Demister Dengan Steam Quality NCG Menggunakan Metode Long Short Term Memory (LSTM). Other thesis, Institut Teknologi Sepuluh Nopember.

Text 2040221144-Undergraduate_thesis.pdf - Accepted Version Restricted to Repository staff only Download (2MB) | Request a copy

Abstract

Kualitas uap (steam quality) merupakan faktor penting yang memengaruhi efisiensi dan keandalan turbin pada Pembangkit Listrik Tenaga Panas Bumi (PLTP). Peningkatan kandungan Non-Condensable Gas (NCG) dapat menurunkan kualitas uap, meningkatkan backpressure, serta mempercepat degradasi komponen turbin. Di PLTP Dieng, peningkatan kandungan NCG menyebabkan penurunan performa turbin sehingga diperlukan sistem yang mampu memberikan peringatan dini sebelum kondisi tersebut mencapai batas kritis. Namun, sistem monitoring yang tersedia saat ini masih berfokus pada pemantauan kondisi secara real-time dan belum mampu memprediksi perubahan kandungan NCG sebagai dasar pengambilan keputusan pemeliharaan. Penelitian ini bertujuan mengembangkan sistem Early Warning Steam Quality berbasis metode Long Short-Term Memory (LSTM) untuk memprediksi kandungan NCG dan mendukung penerapan predictive maintenance. Penelitian menggunakan 1.187 data historis NCG yang diperoleh dari sistem Distributed Control System (DCS) PLTP dengan interval pencatatan 2 jam. Tahapan penelitian meliputi akuisisi data, preprocessing, pembentukan sequence data, pelatihan model LSTM, serta prediksi kandungan NCG. Hasil prediksi selanjutnya digunakan untuk menentukan status Early Warning berdasarkan batas aman kandungan NCG sebesar 1,5% dan mengestimasi Remaining Useful Life (RUL) turbin. Hasil penelitian menunjukkan bahwa model LSTM mampu mempelajari pola temporal data historis NCG dan menghasilkan prediksi dengan nilai Mean Absolute Error (MAE) sebesar 0,14 dan Root Mean Square Error (RMSE) sebesar 0,20, yang menunjukkan tingkat kesalahan prediksi relatif rendah. Sistem yang dikembangkan mampu mendeteksi kecenderungan peningkatan kandungan NCG sebelum mencapai batas aman operasi, menampilkan hasil prediksi, estimasi RUL, dan status Early Warning pada dashboard monitoring. Sistem ini diharapkan dapat mendukung penerapan 0predictive maintenance, meningkatkan keandalan operasi turbin, serta meminimalkan potensi gangguan akibat penurunan kualitas steam.
==================================================================================================================================
Steam quality is a critical factor affecting the efficiency and reliability of turbines in Geothermal Power Plants (GPPs). An increase in the concentration of Non-Condensable Gas (NCG) can reduce steam quality, increase backpressure, and accelerate turbine component degradation. At the Dieng Geothermal Power Plant, elevated NCG levels have led to decreased turbine performance, highlighting the need for an early warning system capable of detecting potential failures before critical operating conditions are reached. However, the existing monitoring system is limited to real-time condition monitoring and is not capable of predicting changes in NCG concentration to support maintenance decision-making. This study aims to develop an Early Warning Steam Quality system based on the Long Short-Term Memory (LSTM) method to predict NCG concentration and support predictive maintenance. The study utilized 1,187 historical NCG data points obtained from the Distributed Control System (DCS) of the geothermal power plant, with a recording interval of two hours. The research stages included data acquisition, preprocessing, sequence generation, LSTM model training, and NCG prediction. The prediction results were then used to determine the Early Warning status based on the safe NCG threshold of 1.5% and to estimate the turbine's Remaining Useful Life (RUL). The results indicate that the LSTM model successfully learned the temporal patterns of historical NCG data and achieved a Mean Absolute Error (MAE) of 0.14 and a Root Mean Square Error (RMSE) of 0.20, indicating a relatively low prediction error. The developed system is capable of detecting increasing trends in NCG concentration before reaching the operational safety threshold and presenting prediction results, RUL estimation, and Early Warning status through a monitoring dashboard. The proposed system is expected to support predictive maintenance, improve turbine operational reliability, and minimize the risk of failures caused by declining steam quality.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Early Warning System, Steam Quality, Non-Condensable Gas (NCG), Long Short-Term Memory (LSTM), Remaining Useful Life (RUL). . Early Warning System, Steam Quality, Non-Condensable Gas (NCG), Long Short-Term Memory (LSTM), Remaining Useful Life (RUL).
Subjects:	A General Works > AI Indexes (General) A General Works > AI Indexes (General) T Technology > T Technology (General) > T57.5 Data Processing T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK1318 Geothermal Power Plants T Technology > TS Manufactures > TS174 Maintainability (Engineering) . Reliability (Engineering)
Divisions:	Faculty of Vocational > 36304-Automation Electronic Engineering
Depositing User:	Rosma Aprilia Dewanti
Date Deposited:	11 Aug 2026 03:02
Last Modified:	11 Aug 2026 03:02
URI:	http://repository.its.ac.id/id/eprint/144296

Actions (login required)

View Item