Analisis Keandalan Photovoltaic Pembangkit Listrik Tenaga Surya di Bantul, Yogyakarta menggunakan Machine Learning

Fitrianingtyas, Patricia (2026) Analisis Keandalan Photovoltaic Pembangkit Listrik Tenaga Surya di Bantul, Yogyakarta menggunakan Machine Learning. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6009211016-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (2MB) | Request a copy

Abstract

Pembangkit Listrik Tenaga Surya merupakan salah satu pengembangan teknologi yang memanfaatkan tenaga surya sebagai salah satu bentuk energi terbarukan untuk mengurangi emisi karbon di atmosfer. Rentan terjadinya kegagalan seperti fault pada sistem PLTS yang dikarenakan pengaruh cuaca, panel photovoltaic yang kotor, adanya bayangan yang menutupi panel, kerusakan pada kabel, short-circuit, kerusakan mekanis dan lain sebagainya. Untuk menyelesaikan permasalahan tersebut diperlukan condition monitoring berbasis real time dengan menggunakan machine learning. Pembangkit Listrik Tenaga Surya Pandansimo, Bantul, Yogyakarta merupakan salah satu PLTS di Indonesia, hasil produksi energi digunakan sebagai sumber energi listrik warga sekitar sebagai pendukung kebutuhan sehari-hari dan UMKM. Nilai reliability, availability dan maintainability sistem photovoltaic sangat mempengaruhi kinerja dalam memberikan suplai tenaga listrik terhadap masyarakat. Model Random Forest Regression digunakan untuk memprediksi nilai tegangan sistem photovoltaic, dengan hasil koefisien determinasi (R²) sebesar 0.5964 yang menunjukkan kemampuan model dalam menjelaskan sekitar 59.74% variasi data aktual. Deteksi kegagalan dilakukan menggunakan metode Residual Based Anomaly Detection (RBAD) dengan batas kontrol 5σ pada data residual, sehingga diperoleh 16 kejadian kegagalan selama 9.664 jam operasional. Berdasarkan hasil tersebut, nilai Mean Time to Failure (MTTF) sistem photovoltaic diperoleh sebesar 593.75 jam dengan laju kegagalan konstan sebesar 0.00168. Analisis reliability menunjukkan penurunan keandalan sistem seiring waktu operasi, di mana nilai reliability mencapai 70% pada 211 jam dan 60% pada 296 jam operasional, yang mengindikasikan perlunya perawatan berkala. Analisis maintainability menghasilkan nilai Mean Time to Repair (MTTR) sebesar 1.0725 jam, dengan nilai maintainability meningkat cepat dan mendekati 1 setelah 11 jam operasional, menunjukkan kemampuan sistem untuk dipulihkan dalam waktu singkat. Analisis availability dengan skenario perawatan rutin setiap 211 jam selama 3 jam menunjukkan bahwa sistem memiliki tingkat ketersediaan yang tinggi, yaitu 98.2% pada 500 jam operasional. Downtime yang relatif singkat dapat dicapai melalui strategi pemeliharaan yang tepat, seperti pembersihan modul photovoltaic sehingga kinerja sistem tetap optimal.
======================================================================================================================================
Photovoltaic power plants represent one of the technological developments that utilize solar energy as a form of renewable energy to reduce carbon emissions at atmosphere. However, photovoltaic systems are susceptible to various types of failures, such as system faults caused by weather conditions, soiled photovoltaic panels, partial shading, cable damage, short circuits, mechanical failures, and other disturbances. To address these issues, real-time condition monitoring based on machine learning is required. Pandansimo Photovoltaic Power Plant, located in Bantul, Yogyakarta, Indonesia, is one of solar power plants that supplies electricity to local communities to support daily activities and UMKM. Reliability, availability, and maintainability of photovoltaic system play a critical role to ensure a stable electricity supply. In this study, a Random Forest Regression model was employed to predict the photovoltaic system voltage, achieving a coefficient of determination (R²) of 0.5964, indicating that the model explains approximately 59.74% of the variance in the actual data. Failure detection was performed using Residual-Based Anomaly Detection (RBAD) with a 5σ control limit applied to the residual data, resulting in 16 detected failure events over 9,664 operating hours. Based on these results, the Mean Time to Failure (MTTF) of the photovoltaic system was calculated as 593.75 hours, corresponding to a constant failure rate of 0.00168. Reliability analysis showed a gradual degradation of system reliability over time, with reliability values decreasing to 70% at 211 operating hours and 60% at 296 operating hours, indicating the necessity of periodic maintenance. Maintainability analysis yielded a Mean Time to Repair (MTTR) of 1.0725 hours, with maintainability increasing rapidly and approaching unity after 11 operating hours, demonstrating the system’s ability to be restored to normal operating conditions within a short time. Furthermore, availability analysis under a preventive maintenance scenario conducted every 211 hours with a repair duration of 3 hours showed a high system availability of 98.2% over 500 operating hours. This results indicate that although system reliability degrades over time, relatively short downtime can be achieved through appropriate maintenance strategies, such as regular cleaning of photovoltaic modules, thereby maintaining optimal system performance.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	energi terbarukan, Photovoltaic, reliability, predictive maintenance, machine learning, random forest renewable energy, Photovoltaic, reliability, predictive maintenance, machine learning, random forest regression
Subjects:	T Technology > TA Engineering (General). Civil engineering (General) > TA169 Reliability (Engineering)
Divisions:	Faculty of Industrial Technology and Systems Engineering (INDSYS) > Physics Engineering > 30101-(S2) Master Thesis
Depositing User:	Patricia Fitrianingtyas
Date Deposited:	13 Feb 2026 01:08
Last Modified:	13 Feb 2026 01:08
URI:	http://repository.its.ac.id/id/eprint/132429

Actions (login required)

View Item