Klasifikasi Kondisi Partial Shading pada Sistem Photovoltaic Kapasitas 2,2 kW Berdasarkan Karakteristik Kurva I-V dan P-V Menggunakan Metode Support Vector Machine (SVM)

Aulia, Rahma Nadia (2026) Klasifikasi Kondisi Partial Shading pada Sistem Photovoltaic Kapasitas 2,2 kW Berdasarkan Karakteristik Kurva I-V dan P-V Menggunakan Metode Support Vector Machine (SVM). Other thesis, Institut Teknologi Sepuluh Nopember.

Text 2040221085-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (6MB) | Request a copy

Abstract

Partial shading merupakan kondisi ketika sebagian permukaan panel surya menerima iradiansi yang lebih rendah dibandingkan bagian lainnya sehingga menyebabkan perubahan karakteristik tegangan, arus, dan daya. Perubahan tersebut memengaruhi bentuk kurva arus-tegangan (I-V) dan daya-tegangan (P-V), sehingga identifikasi tingkat partial shading secara langsung menjadi sulit ketika jumlah dan variasi data semakin besar. Penelitian ini bertujuan mengembangkan model klasifikasi kondisi partial shading untuk konteks sistem photovoltaic berkapasitas 2,2 kW. Pengujian karakteristik dan pembentukan dataset dilakukan menggunakan panel surya 120 Wp sebagai media eksperimen yang lebih aman serta memungkinkan variasi beban dikendalikan. Kondisi pengujian terdiri atas kelas K0 atau non-shading serta kelas K1 sampai K5 yang mewakili tingkat shading 15%, 30%, 45%, 60%, dan 75%, dengan dua pola penutupan pada setiap tingkat shading. Tegangan dan arus diukur menggunakan PZEM-017, sedangkan daya diperoleh dari hasil perkalian tegangan dan arus. Sebanyak 14.070 baris data diproses melalui tahap pelabelan, forward fill, cleaning, segmentasi menjadi 84 sesi, dan pembentukan 924 window berukuran 15 titik. Dari setiap window diekstraksi 29 fitur yang digunakan sebagai masukan model Support Vector Machine. Model terbaik menggunakan kernel linear dengan parameter C sebesar 10. Hasil pengujian menunjukkan accuracy per-window sebesar 87,45%, macro F1-score sebesar 0,8581, dan accuracy per-sesi sebesar 100% melalui majority vote pada 21 sesi testing. Hasil tersebut menunjukkan bahwa kombinasi fitur tegangan, arus, dan daya mampu membedakan enam kelas kondisi shading pada dataset penelitian. Namun, penerapan model secara langsung pada sistem 2,2 kW tetap memerlukan pelatihan ulang menggunakan data dari konfigurasi empat modul 550 Wp agar model sesuai dengan rentang dan karakteristik sistem target.
====================================================================================================================================
Partial shading occurs when different parts of a photovoltaic module receive unequal irradiance, causing changes in voltage, current, and power characteristics. These changes affect the current-voltage (I-V) and power-voltage (P-V) curves, making direct identification of partial-shading levels difficult when the amount and variation of data increase. This study aims to develop a partial-shading classification model for the context of a 2.2 kW photovoltaic system. Characteristic testing and dataset construction were conducted using a 120 Wp photovoltaic module as a safer experimental platform that allowed controlled load variation. The test conditions consisted of K0 or non-shading and K1 to K5 representing 15%, 30%, 45%, 60%, and 75% shading, with two covering patterns at each shading level. Voltage and current were measured using a PZEM-017, while power was calculated by multiplying voltage and current. A total of 14,070 data rows were processed through labeling, forward filling, cleaning, segmentation into 84 sessions, and formation of 924 windows containing 15 consecutive points. Twenty-nine features were extracted from each window and used as inputs to the Support Vector Machine model. The best model used a linear kernel with C = 10. The testing results showed a per-window accuracy of 87.45%, a macro F1-score of 0.8581, and a session-level accuracy of 100% using majority voting on 21 testing sessions. The results indicate that combined voltage, current, and power features can distinguish six shading classes within the research dataset. However, direct implementation on the 2.2 kW system still requires retraining using data from the actual four-module 550 Wp configuration so that the model matches the target system range and characteristics.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Photovoltaic, Partial Shading, Kurva I-V, Kurva P-V, Support Vector Machine
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines.
Divisions:	Faculty of Vocational > 36304-Automation Electronic Engineering
Depositing User:	Rahma Nadia Aulia
Date Deposited:	10 Aug 2026 01:50
Last Modified:	10 Aug 2026 01:50
URI:	http://repository.its.ac.id/id/eprint/142121

Actions (login required)

View Item