Model Optimasi Prediksi Aspek Ekonomis Dan Operasional Pada Sistem Water Treatment Plant (WTP) Pembangkit Listrik Tenaga Uap (PLTU)

Hamdani, Hamdani (2026) Model Optimasi Prediksi Aspek Ekonomis Dan Operasional Pada Sistem Water Treatment Plant (WTP) Pembangkit Listrik Tenaga Uap (PLTU). Doctoral thesis, Institut Teknologi Sepuluh Nopember.

Text 7032212002-Master_Thsis.pdf - Accepted Version Restricted to Repository staff only Download (3MB) | Request a copy

Abstract

Keandalan Pembangkit Listrik Tenaga Uap (PLTU) sangat dipengaruhi oleh kinerja Water Treatment Plant (WTP) dalam menghasilkan air demineralisasi yang memenuhi spesifikasi operasi boiler dan turbin. Selain menentukan keandalan operasi, kinerja WTP juga berkontribusi terhadap biaya operasi dan biaya pemeliharaan pembangkit. Namun demikian, penelitian terdahulu umumnya masih berfokus pada prediksi parameter operasi Reverse Osmosis (RO) atau menggunakan data laboratorium, sehingga belum mampu mengintegrasikan aspek operasional dan ekonomis berbasis data aktual industri dalam suatu model prediksi yang teroptimasi. Kesenjangan tersebut menunjukkan perlunya pengembangan model yang mampu merepresentasikan hubungan nonlinier antarvariabel operasi sekaligus menghasilkan prediksi yang akurat sebagai dasar pengambilan keputusan. Penelitian ini bertujuan mengembangkan model prediksi terintegrasi aspek operasional dan ekonomis pada Water Treatment Plant berbasis Reverse Osmosis menggunakan Artificial Neural Network (ANN) yang dioptimasi dengan Genetic Algorithm (GA). Model dikembangkan menggunakan data historis operasi WTP PLTU selama dua tahun. Sebelum pemodelan, dilakukan seleksi variabel menggunakan Expert Judgment dan Random Forest untuk mengidentifikasi parameter yang paling berpengaruh terhadap keluaran sistem. Model ANN selanjutnya dioptimasi menggunakan GA untuk memperoleh konfigurasi jaringan terbaik dalam memprediksi biaya operasi, biaya pemeliharaan, serta parameter kualitas air demineralisasi. Hasil penelitian menunjukkan bahwa integrasi Random Forest–ANN–GA mampu meningkatkan akurasi prediksi secara signifikan dibandingkan ANN tanpa optimasi. Nilai Mean Absolute Percentage Error (MAPE) pada aspek ekonomis menurun dari 11,68% menjadi 1,76%, sedangkan pada aspek operasional menurun dari 3,91% menjadi 2,65%. Hasil tersebut menunjukkan bahwa optimasi menggunakan GA efektif meningkatkan kemampuan ANN dalam memodelkan sistem WTP yang kompleks dan nonlinier. Kebaruan penelitian ini terletak pada pengembangan model prediksi terintegrasi yang menggabungkan seleksi variabel menggunakan Random Forest, pemodelan nonlinier menggunakan ANN, dan optimasi hyperparameter menggunakan GA dengan memanfaatkan data historis industri aktual. Model yang dihasilkan tidak hanya meningkatkan akurasi prediksi, tetapi juga dapat diterapkan sebagai dasar pengambilan keputusan untuk mendukung predictive operation, predictive maintenance, dan cost optimization, sehingga berkontribusi terhadap peningkatan efisiensi biaya, kualitas air demineralisasi, dan keandalan operasi PLTU.
========================================================================================================================================
The reliability of Steam Power Plants (PLTU) is highly dependent on the performance of the Water Treatment Plant (WTP) in producing demineralized water that meets the operational requirements of boilers and turbines. In addition to ensuring operational reliability, WTP performance significantly influences both operational and maintenance costs of power plants. However, previous studies have primarily focused on predicting the operational performance of Reverse Osmosis (RO) systems or have relied on laboratory-scale data, with limited attention to integrating both operational and economic aspects using actual industrial data within an optimized predictive framework. This research gap highlights the need for a predictive model capable of representing the complex nonlinear relationships among operational variables while providing accurate predictions to support decision-making. This study aims to develop an integrated predictive model for the operational and economic aspects of a Reverse Osmosis-based Water Treatment Plant using an Artificial Neural Network (ANN) optimized by a Genetic Algorithm (GA). The model was developed using two years of historical operational data collected from a utility-scale Steam Power Plant WTP. Prior to model development, variable selection was conducted through expert judgment and the Random Forest algorithm to identify the most influential input parameters affecting system performance. The selected variables were subsequently used to develop the ANN model, which was further optimized using GA to determine the optimal network architecture for predicting operational costs, maintenance costs, and key demineralized water quality parameters. The results demonstrate that the integrated Random Forest–ANN–GA approach substantially improves predictive performance compared with the conventional ANN model. The Mean Absolute Percentage Error (MAPE) for the economic model decreased from 11.68% to 1.76%, while the MAPE for the operational model decreased from 3.91% to 2.65%. These findings indicate that GA-based optimization effectively enhances the predictive capability of ANN in modeling the complex and nonlinear behavior of WTP systems. The novelty of this research lies in the development of an integrated predictive framework that combines expert judgment, Random Forest-based variable selection, ANN-based nonlinear modeling, and GA-based hyperparameter optimization using actual industrial historical data. The proposed model not only improves prediction accuracy but also provides a practical foundation for decision-making by supporting predictive operation, predictive maintenance, and cost optimization. Consequently, it contributes to improving cost efficiency, demineralized water quality, and the operational reliability of Steam Power Plants.

Item Type:	Thesis (Doctoral)
Uncontrolled Keywords:	Water Treatment Plant (WTP), Reverse Osmosis, Artificial Neural Network (ANN), Genetic Algorithm (GA), Random Forest, model prediksi terintegrasi
Subjects:	Q Science > QA Mathematics > QA402.5 Genetic algorithms. Interior-point methods. T Technology > TD Environmental technology. Sanitary engineering > TD433 Water treatment plants
Divisions:	Interdisciplinary School of Management and Technology (SIMT) > 61101-Master of Technology Management (MMT)
Depositing User:	Hamdani Hamdani
Date Deposited:	27 Jul 2026 05:10
Last Modified:	27 Jul 2026 05:10
URI:	http://repository.its.ac.id/id/eprint/138885

Actions (login required)

View Item