Analisis Peramalan Output Daya Pembangkit Listrik Tenaga Surya (PLTS) dengan Pemodelan Hybrid Fungsi Transfer Single Output-Single Input dan LSTM

Sa'adah, Yunia (2026) Analisis Peramalan Output Daya Pembangkit Listrik Tenaga Surya (PLTS) dengan Pemodelan Hybrid Fungsi Transfer Single Output-Single Input dan LSTM. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6003232019-Master_Thesis.pdf Restricted to Repository staff only Download (3MB) | Request a copy

Abstract

Kebutuhan energi listrik di Indonesia yang terus meningkat menuntut pemanfaatan sumber energi terbarukan yang lebih optimal, termasuk Pembangkit Listrik Tenaga Surya (PLTS). Di wilayah Z, potensi energi surya cukup tinggi, namun variabilitas cuaca menyebabkan fluktuasi output daya sehingga diperlukan metode peramalan yang akurat. Penelitian ini mengembangkan model hybrid berbasis fungsi transfer, Generalized Autoregressive Conditional Heteroskedasticity (GARCH), dan Long Short-Term Memory (LSTM) untuk meningkatkan akurasi peramalan output daya PLTS. Pendekatan hybrid dipilih karena mampu menggabungkan keunggulan model linear dalam menangkap hubungan antar variabel input dan pola musiman, dengan kemampuan GARCH dalam memodelkan dinamika volatilitas error serta kemampuan LSTM dalam menangkap pola non-linear dan dependensi jangka panjang pada data deret waktu. Model fungsi transfer terbaik untuk memodelkan output daya listrik PLTS wilayah Z adalah (b=2,r=1,s=0)(2,0,0)(0,0,1)23. Model fungsi transfer output daya listrik pada suatu periode dipengaruhi oleh output daya listrik satu, dua dan tiga periode sebelumnya serta dipengaruhi oleh irradiance dua, tiga dan empat periode sebelumnya. Selain itu, output daya listrik pada suatu periode juga dipengaruhi oleh komponen error pada suatu satu, 23 dan 24 periode sebelumnya. Hasil evaluasi menunjukkan bahwa model fungsi transfer tunggal menghasilkan RMSE sebesar 185,8688. Setelah residual dimodelkan menggunakan GARCH, kinerja peramalan meningkat dengan RMSE sebesar 149,6496. Selanjutnya, pemodelan residual hybrid fungsi transfer–GARCH menggunakan LSTM menghasilkan peningkatan akurasi lebih lanjut dengan RMSE sebesar 149,0787. Secara keseluruhan, model hybrid fungsi transfer–GARCH–LSTM mampu mengoreksi kesalahan prediksi sebesar 36,7901 dibandingkan model fungsi transfer tunggal. Hasil ini menunjukkan bahwa pendekatan hybrid mampu meningkatkan keandalan peramalan daya listrik PLTS dan berpotensi mendukung perencanaan serta pengelolaan sistem tenaga surya secara lebih optimal di wilayah Z.
================================================================================================================================
The increasing demand for electrical energy in Indonesia requires more optimal utilization of renewable energy sources, including Solar Power Plants (PLTS). In region Z, solar energy potential is relatively high; however, weather variability causes fluctuations in power output, making accurate forecasting methods essential. This study develops a hybrid model based on transfer function, Generalized Autoregressive Conditional Heteroskedasticity (GARCH), and Long Short-Term Memory (LSTM) to improve the accuracy of PLTS power output forecasting. The hybrid approach is selected because it combines the advantages of linear models in capturing relationships among input variables and seasonal patterns, with the capability of GARCH in modeling error volatility dynamics, as well as the ability of LSTM to capture non-linear patterns and long-term dependencies in time series data. The best transfer function model is (b=2,r=1,s=0)(2,0,0)(0,0,1)23. The transfer function model indicates that power output at a given period is influenced by power output from one, two, and three previous periods, as well as by irradiance from two, three, and four previous periods. In addition, power output at a given period is also affected by the error components from one, 23, and 24 previous periods. The evaluation results show that the single transfer function model produces an RMSE of 185.8688. After the residuals were modeled using GARCH, the forecasting performance improved with an RMSE of 149.6496. Furthermore, modeling the residuals of the transfer function–GARCH hybrid using LSTM resulted in further accuracy improvement with an RMSE of 149.0787. Overall, the hybrid transfer function–GARCH–LSTM model is able to reduce the prediction error by 36.7901 compared to the single transfer function model. These results indicate that the hybrid approach can improve the reliability of PLTS power output forecasting and has the potential to support planning and operational management of solar power systems more optimally in region Z.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	PLTS, Fungsi Transfer, GARCH, LSTM, Hybrid, PLTS, Transfer Function,GARCH, LSTM, Hybrid
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. Q Science > QA Mathematics > QA280 Box-Jenkins forecasting
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49101-(S2) Master Thesis
Depositing User:	Yunia Sa'adah Sa'adah
Date Deposited:	04 Feb 2026 08:59
Last Modified:	04 Feb 2026 08:59
URI:	http://repository.its.ac.id/id/eprint/132134

Actions (login required)

View Item