Model Prediktif Emisi Karbon Dari PLTU Berbahan Bakar Batubara Untuk Membantu Pengambilan Keputusan Perdagangan Karbon

Ismail, Mahsun (2026) Model Prediktif Emisi Karbon Dari PLTU Berbahan Bakar Batubara Untuk Membantu Pengambilan Keputusan Perdagangan Karbon. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6032241166-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (4MB) | Request a copy

Abstract

Perubahan iklim akibat emisi gas rumah kaca (GRK) menjadi tantangan global terutama dari sektor energi khususnya pada sub sektor pembangkit tenaga listrik yang bergantung pada pembangkit listrik tenaga uap (PLTU) batubara. Pemerintah Indonesia menargetkan penurunan emisi melalui Nilai Ekonomi Karbon (NEK) dan mekanisme perdagangan karbon. Penelitian ini mengembangkan model prediktif berbasis Machine Learning untuk memprediksi emisi CO₂ dari operasional PLTU batubara guna mendukung pengambilan keputusan strategis dalam perdagangan karbon. Data operasional PLTU XYZ Unit Y berkapasitas 660 MW (net) periode Januari-Desember 2025 sebanyak 14.549 baris dianalisis menggunakan Artificial Neural Network (ANN), Extreme Gradient Boosting (XGBoost) dan Support Vector Regression (SVR) dengan metrik evaluasi RMSE, MAE dan R². XGBoost memberikan performa terbaik (MAE 15,84 ton/jam (2,56%); RMSE 20,53 ton/jam (3,32%); R² pengujian 0,956) dan ditetapkan sebagai model utama. Parameter paling berpengaruh terhadap emisi CO₂ adalah COAL_FLOW (Feature Importance 0,588), O2_BOILER (0,235), dan FEGT_AVG (0,072). Optimasi dua parameter terkendali (O2_BOILER, FEGT_AVG) dengan skenario Pesimis (P20-P80), Moderat (P10-P90), dan Optimis (P5-P95) menghasilkan penurunan intensitas emisi 2,33% - 3,49% dengan reduksi absolut 104.155 - 155.804 t-CO₂ pada proyeksi 2026. Di skenario Moderat, optimasi dapat menurunkan intensitas emisi 2,64% dari kondisi BAU setara 117.888 t-CO₂ sehingga meningkatkan surplus Unit Y dari 136.205 menjadi 254.093 t-CO₂ terhadap estimasi kuota emisi 4.599.623,10 t-CO₂. Pada harga referensi IDXCarbon Rp 58.800 per ton CO₂, pendapatan meningkat dari Rp 8,01 miliar (BAU) menjadi Rp 14,13 - 17,17 miliar setelah optimasi, atau Rp 14,94 miliar pada skenario Moderat dengan tambahan manfaat ekonomi Rp 6,12 - 9,16 miliar (Rp 6,93 miliar pada skenario Moderat). Temuan ini menunjukkan bahwa model prediktif berbasis Machine Learning mampu memprediksi emisi CO₂ secara akurat dan melalui optimasi parameter operasional berkontribusi nyata terhadap pengendalian emisi dan peningkatan nilai ekonomi pembangkit dalam mekanisme perdagangan karbon.
========================================================================================================================================
Climate change driven by rising greenhouse gas (GHG) emissions poses a critical global challenge particularly in the energy sector especially in the power generation sub-sector that still relies on coal-fired steam power plants (PLTU). Indonesia targets an emission reduction through the Carbon Economic Value (NEK) framework and carbon trading mechanisms. This study develops a Machine Learning-based predictive model to estimate CO₂ emissions from coal-fired power plant operations and support strategic carbon trading decisions. Operational data from PLTU XYZ Unit Y (660 MW net capacity) covering January–December 2025 totalling 14,549 records were analysed using Artificial Neural Network (ANN), Extreme Gradient Boosting (XGBoost) and Support Vector Regression (SVR) with evaluation metrics RMSE, MAE and R². XGBoost achieved the best performance (MAE 15.84 ton/h (2.56%); RMSE 20.53 ton/h (3.32%); R² test 0.956) and was selected as the primary model. The most influential parameters on CO₂ emissions were COAL_FLOW (Feature Importance 0.588), O2_BOILER (0.235), and FEGT_AVG (0.072). Optimization of two controllable parameters (O2_BOILER, FEGT_AVG) under Pessimistic (P20-P80), Moderate (P10-P90), and Optimistic (P5-P95) scenarios yielded emission intensity reductions of 2.33% - 3.49% with absolute reductions of 104,155 - 155,804 t-CO₂ in the 2026 projection. Under the Moderate scenario, optimization reduced emission intensity by 2.64% from BAU equivalent to 117,888 t-CO₂ increasing Unit Y's surplus from 136,205 to 254,093 t-CO₂ against the estimated emission quota allocation of 4,599,623.10 t-CO₂. At the IDXCarbon reference price of IDR 58,800 per ton CO₂, revenue increased from IDR 8.01 billion (BAU) to IDR 14.13 - 17.17 billion after optimization, or IDR 14.94 billion under the Moderate scenario, with additional economic benefits of IDR 6.12 - 9.16 billion (IDR 6.93 billion under the Moderate scenario). These findings demonstrate that the Machine Learning-based predictive model accurately predicts CO₂ emissions and through operational parameter optimization delivers tangible contributions to emission control and economic value improvement within the carbon trading mechanism.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Prediksi emisi, Emisi karbon, PLTU batubara, Machine Learning, Perdagangan karbon, Emission prediction, Carbon emissions, Coal-fired power plant, Machine Learning, Carbon trading
Subjects:	T Technology > TD Environmental technology. Sanitary engineering > TD171.75 Climate change mitigation
Divisions:	Interdisciplinary School of Management and Technology (SIMT) > 61101-Master of Technology Management (MMT)
Depositing User:	Mahsun Ismail
Date Deposited:	23 Jul 2026 10:24
Last Modified:	23 Jul 2026 10:24
URI:	http://repository.its.ac.id/id/eprint/137253

Actions (login required)

View Item