Prediksi Daya Listrik pada Pembangkit Listrik Tenaga Panas Bumi Menggunakan Pendekatan Machine Learning dan Explainable Artificial Intelligence

Haryadi, Nur (2026) Prediksi Daya Listrik pada Pembangkit Listrik Tenaga Panas Bumi Menggunakan Pendekatan Machine Learning dan Explainable Artificial Intelligence. Masters thesis, Institut Teknologi Sepuluh Nopember.

This is the latest version of this item.

[thumbnail of 6025241003-Master_Thesis.pdf] Text
6025241003-Master_Thesis.pdf - Accepted Version
Restricted to Repository staff only
Download (5MB)

Abstract

Pembangkit Listrik Tenaga Panas Bumi (PLTP) merupakan salah satu sumber energi terbarukan yang berperan penting dalam mendukung ketahanan energi nasional. Pada PLTP PT Geo Dipa Energi (Persero) Unit Dieng, sistem operasional telah mampu melakukan pemantauan data secara real-time, namun belum memiliki kemampuan prediksi daya listrik berbasis Machine Learning sehingga proses antisipasi terhadap perubahan daya masih bergantung pada analisis data historis dan pengalaman operator. Selain itu, hubungan antara parameter operasional dan daya listrik bersifat kompleks serta non-linier sehingga diperlukan pendekatan prediksi yang akurat dan dapat diinterpretasikan. Penelitian ini menggunakan data operasional PLTP PT Geo Dipa Energi (Persero) Unit Dieng periode 2022–2024 untuk mengembangkan model prediksi daya listrik menggunakan lima algoritma Machine Learning, yaitu Decision Tree Regression (DTR), eXtreme Gradient Boosting (XGB), Extremely Randomized Trees (XRT), Natural Gradient Boosting (NGB), dan Deep Neural Network (DNN). Optimasi hyperparameter dilakukan menggunakan Randomized Search Cross Validation, sedangkan evaluasi model menggunakan Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), dan koefisien determinasi (R²). Interpretabilitas model dianalisis melalui pendekatan Explainable Artificial Intelligence (XAI) menggunakan metode SHAP dan LIME pada model terbaik. Hasil penelitian menunjukkan bahwa model XRT memberikan performa terbaik dengan nilai RMSE sebesar 0.05190, MAE sebesar 0.03338, dan R² sebesar 0.99994. Model DTR, XGB, DNN, dan NGB memperoleh nilai R² berturut-turut sebesar 0.99844, 0.99376, 0.99000, dan 0.95588. Analisis SHAP dan LIME menunjukkan bahwa fitur-fitur operasional utama memberikan kontribusi dominan terhadap pembangkitan daya listrik serta mampu menjelaskan hasil prediksi secara global dan lokal.
================================================================================================================================
Geothermal Power Plants (GPPs) play an important role in supporting national energy security as one of the most reliable renewable energy sources. At PT Geo Dipa Energi (Persero) Dieng Unit, the operational system is capable of monitoring plant data in real time; however, it has not yet implemented Machine Learning-based electricity generation prediction. Consequently, anticipation of power generation changes still relies on historical data analysis and operator experience. Furthermore, the relationship between operational parameters and electricity generation is complex and nonlinear, requiring a predictive approach that is both accurate and interpretable. This study utilizes operational data collected from PT Geo Dipa Energi (Persero) Dieng Unit during the period of 2022-2024 to develop an electricity generation prediction model using five Machine Learning algorithms, namely Decision Tree Regression (DTR), eXtreme Gradient Boosting (XGB), Extremely Randomized Trees (XRT), Natural Gradient Boosting (NGB), and Deep Neural Network (DNN). Hyperparameter optimization was performed using Randomized Search Cross Validation, while model performance was evaluated using Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and the coefficient of determination (R²). Model interpretability was investigated using the Explainable Artificial Intelligence (XAI) approach through SHAP and LIME on the best-performing model. The results demonstrate that the XRT algorithm achieved the best predictive performance with an RMSE of 0.05190, an MAE of 0.03338, and an R² of 0.99994. Meanwhile, the DTR, XGB, DNN, and NGB models achieved R² values of 0.99844, 0.99376, 0.99000, and 0.95588, respectively. SHAP and LIME analyses revealed that the key operational features contributed dominantly to electricity generation prediction and successfully provided both global and local explanations of the model outputs. These findings indicate that integrating Machine Learning with XAI can produce an accurate, reliable, and interpretable prediction model, thereby supporting operational decision-making in geothermal power plants.

Item Type: Thesis (Masters)
Uncontrolled Keywords: Explainable AI, LIME, Machine Learning, PLTP, Prediksi Daya Listrik, SHAP Electricity Power Prediction, Explainable AI, Geothermal Power Plant, LIME, Machine Learning, SHAP
Subjects: T Technology > T Technology (General) > T57.5 Data Processing
T Technology > T Technology (General) > T57.8 Nonlinear programming. Support vector machine. Wavelets. Hidden Markov models.
T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK1001 Production of electric energy or power
T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK1318 Geothermal Power Plants
Divisions: Faculty of Electrical Technology > Electrical Engineering > 20101-(S2) Master Thesis
Depositing User: Nur Haryadi
Date Deposited: 30 Jul 2026 03:08
Last Modified: 30 Jul 2026 03:08
URI: http://repository.its.ac.id/id/eprint/139628

Available Versions of this Item

Actions (login required)

View Item View Item