Pengembangan Predictive Analytics Berbasis Machine Learning untuk Memprediksi Parameter Suhu Transformator di PT Maxima Daya Indonesia

Endrian, Herbert (2026) Pengembangan Predictive Analytics Berbasis Machine Learning untuk Memprediksi Parameter Suhu Transformator di PT Maxima Daya Indonesia. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5010221178-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (8MB) | Request a copy

Abstract

Transformator distribusi merupakan komponen penting dalam sistem tenaga listrik karena berperan dalam menjaga kontinuitas penyaluran energi listrik. Seiring berkembangnya Internet of Things (IoT), data sensor transformator dapat dimanfaatkan untuk mendukung predictive analytics dalam memprediksi kondisi transformator di masa mendatang. Penelitian ini bertujuan membangun dan membandingkan model prediksi suhu transformator menggunakan Naive Forecast sebagai statistical baseline, serta Random Forest (RF), Extreme Gradient Boosting (XGBoost), dan Support Vector Regression (SVR) sebagai model machine learning. Data sensor diproses melalui tahapan feature engineering, kemudian setiap model dibangun pada horizon prediksi 1 jam, 2 jam, 6 jam, 12 jam, dan 24 jam menggunakan hyperparameter awal maupun Bayesian Optimization. Kinerja model dievaluasi menggunakan metrik MAE, RMSE, MAPE, R², WMAE, dan WRMSE. Hasil penelitian menunjukkan bahwa penerapan Bayesian Optimization mampu meningkatkan performa model machine learning, khususnya pada model Random Forest dan XGBoost. Berdasarkan mekanisme hierarchical ranking menggunakan WMAE sebagai metrik utama, model XGBoost dengan Bayesian Optimization memberikan performa terbaik pada horizon prediksi 1 jam, 2 jam, 6 jam, dan 12 jam, sedangkan Random Forest dengan Bayesian Optimization memberikan performa terbaik pada horizon prediksi 24 jam. Selain itu, penelitian ini berhasil mengembangkan dashboard berbasis Python menggunakan Streamlit untuk memvisualisasikan data historis, kondisi aktual, dan hasil prediksi suhu transformator.
========================================================================================================================================
Distribution transformers are critical components of power systems as they ensure the continuity of electrical power distribution. With the advancement of the Internet of Things (IoT), transformer sensor data can be utilized to support predictive analytics for forecasting future transformer conditions. This study aims to develop and compare transformer temperature prediction models using Naive Forecast as the statistical baseline, alongside Random Forest (RF), Extreme Gradient Boosting (XGBoost), and Support Vector Regression (SVR) as machine learning models. The sensor data were processed through a feature engineering stage, after which each model was developed for prediction horizons of 1 hour, 2 hours, 6 hours, 12 hours, and 24 hours using both the initial hyperparameter configuration and Bayesian Optimization. Model performance was evaluated using MAE, RMSE, MAPE, R², WMAE, and WRMSE. The results indicate that Bayesian Optimization improved the performance of the machine learning models, particularly Random Forest and XGBoost. Based on the hierarchical ranking mechanism using WMAE as the primary evaluation metric, XGBoost with Bayesian Optimization achieved the best performance for the 1-hour, 2-hour, 6-hour, and 12-hour prediction horizons, while Random Forest with Bayesian Optimization achieved the best performance for the 24-hour prediction horizon. Furthermore, this study successfully developed a Python-based dashboard using the Streamlit library to visualize historical data, current transformer conditions, and transformer temperature prediction results.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Predictive Analytics, Machine Learning, Suhu Transformator, Internet of Things (IoT), Bayesian Optimization, Predictive Analytics, Machine Learning, Transformer Temperature, Internet of Things (IoT), Bayesian Optimization.
Subjects:	H Social Sciences > HA Statistics > HA30.3 Time-series analysis Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines.
Divisions:	Faculty of Industrial Technology and Systems Engineering (INDSYS) > Industrial Engineering > 26201-(S1) Undergraduate Thesis
Depositing User:	Herbert Endrian
Date Deposited:	03 Aug 2026 07:56
Last Modified:	03 Aug 2026 07:56
URI:	http://repository.its.ac.id/id/eprint/142289

Actions (login required)

View Item