Rachmadani, Risma Okta (2026) Forecasting Daya dan Energi PLTS Off-Grid Kp. Ciputri Berbasis Hybrid Wavelet–LSSVM sebagai Dasar Analisis Kinerja Hybrid Solar Inverter. Other thesis, Institut Teknologi Sepuluh Nopember.
![[thumbnail of 2040221114-Undergraduate_Thesis.pdf]](http://repository.its.ac.id/style/images/fileicons/text.png) |
Text
2040221114-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (9MB) | Request a copy |
Abstract
Kerusakan hybrid solar inverter dan keterbatasan sistem monitoring pada PLTS off-grid Kampung Ciputri menyebabkan belum tersedianya acuan daya dan energi yang seharusnya dibangkitkan berdasarkan kondisi meteorologis dan operasional sistem. Penelitian ini mengembangkan model forecasting berbasis Hybrid Wavelet–Least Squares Support Vector Machine (LSSVM) menggunakan data weather station dan dua hybrid solar inverter yang diintegrasikan pada interval 30 menit dalam rentang operasi PV pukul 06.00–18.00 WIB. Wavelet digunakan untuk denoising, sedangkan LSSVM dengan kernel RBF digunakan untuk memodelkan hubungan nonlinear pada data. Pada next-step forecasting, model menghasilkan MAE 294,68 W, RMSE 412,04 W, dan R^2 0,8882. Pada day-ahead forecasting, diperoleh MAE 294,68 W, RMSE 412,04 W, dan R^2 0,7729, sedangkan forecasting energi menghasilkan MAE 1,27 kWh/hari dan sMAPE 11,13%. Proyeksi langsung selama tujuh hari menghasilkan estimasi energi total 778,95 kWh. Validasi terhadap data aktual 1–7 Mei 2026 menghasilkan RMSE daya 309,94 W dan R^2 0,8640, serta MAE energi 1,01 kWh/hari dan sMAPE 9,21%. Analisis residual menunjukkan 147 dari 154 interval (95,45%) berada pada kondisi normal, 4 dari 154 interval (2,60%) warning, dan 3 dari 154 interval (1,95%) deviasi. Hasil tersebut menunjukkan bahwa model dapat menyediakan acuan expected power dan expected energy serta profil operasi normal untuk mendukung evaluasi kondisi operasi PLTS off-grid di Kampung Ciputri, Cianjur, Jawa Barat.
===================================================================================================================================================
Damage to the hybrid solar inverter and limitations of the monitoring system at the off-grid photovoltaic (PV) system in Kampung Ciputri have resulted in the absence of reliable power and energy references that should be generated under specific meteorological and operational conditions. This study develops a forecasting model based on Hybrid Wavelet–Least Squares Support Vector Machine (LSSVM) using data from a weather station and two hybrid solar inverters, integrated at 30-minute intervals within the PV operating period of 06:00–18:00 Western Indonesian Time (WIB). Wavelet is applied for denoising, while LSSVM with a radial basis function (RBF) kernel is used to model nonlinear relationships within the data. For next-step forecasting, the model achieved an MAE of 294.68 W, an RMSE of 412.04 W, and an R^2 of 0.8882. For day-ahead forecasting, the model achieved an MAE of 294.68 W, an RMSE of 412.04 W, and an R^2 of 0.7729, while energy forecasting resulted in an MAE of 1.27 kWh/day and an sMAPE of 11.13%. The direct seven-day projection produced an estimated total energy generation of 778.95 kWh. Validation against actual data from 1–7 May 2026 resulted in a power RMSE of 309.94 W and an R^2 of 0.8640, as well as an energy MAE of 1.01 kWh/day and an sMAPE of 9.21%. Residual analysis showed that 147 of 154 intervals (95.45%) were classified as normal, 4 of 154 intervals (2.60%) as warnings, and 3 of 154 intervals (1.95%) as deviations. These results indicate that the model can provide expected power and expected energy references, together with a normal operating profile, to support the evaluation of operating conditions of the off-grid PV system in Kampung Ciputri, Cianjur, West Java.
Actions (login required)
 |
View Item |