Penerapan Model Hybrid SARIMA–LSTM dalam Peramalan Global Horizontal Irradiance Bulanan di Kabupaten Purwakarta

Heryanto, Andreas (2026) Penerapan Model Hybrid SARIMA–LSTM dalam Peramalan Global Horizontal Irradiance Bulanan di Kabupaten Purwakarta. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5003221113-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (8MB) | Request a copy

Abstract

Kabupaten Purwakarta memiliki peran strategis dalam pengembangan energi surya nasional dengan beroperasinya PLTS Terapung Cirata dan PLTS Aruna sebagai pembangkit listrik tenaga surya terbesar di Indonesia. Keandalan produksi energi surya dipengaruhi oleh fluktuasi Global Horizontal Irradiance (GHI) yang mengandung pola musiman linier dan komponen nonlinier. Penelitian ini bertujuan untuk mendeskripsikan karakteristik GHI bulanan di Kabupaten Purwakarta serta menerapkan metode Hybrid Seasonal Autoregressive Integrated Moving Average–Long Short-Term Memory (SARIMA–LSTM) untuk menghasilkan peramalan GHI bulanan yang akurat menggunakan data rata-rata GHI harian bulanan periode Januari 1984 hingga Desember 2025 yang bersumber dari NASA Prediction of Worldwide Energy Resources (POWER). Model SARIMA digunakan untuk menangkap pola linier dan musiman, sedangkan LSTM diterapkan pada residual SARIMA untuk memodelkan pola nonlinier yang masih tersisa berdasarkan hasil uji Terasvirta. Model SARIMA terbaik yang diperoleh adalah SARIMA(1,0,1) (0,1,1)12. Selanjutnya, model Hybrid SARIMA–LSTM menghasilkan nilai error yang lebih rendah dibandingkan model SARIMA. Hasil tersebut mengindikasikan bahwa penambahan model LSTM pada residual SARIMA mampu meningkatkan performa peramalan dengan menangkap pola nonlinier residual yang belum dimodelkan secara optimal oleh model linear SARIMA. Hasil peramalan tahun 2026 menunjukkan pola musiman yang konsisten dengan karakteristik historis data, dengan nilai GHI terendah diprediksi terjadi pada Februari 2026 dan tertinggi pada September 2026.
================================================================================================================================
Purwakarta Regency has a strategic role in the development of national solar energy infrastructure with the operation of Cirata Floating Solar Power Plant and Aruna Solar Power Plant as the largest solar power plants in Indonesia. The reliability of solar energy production is influenced by fluctuations in Global Horizontal Irradiance (GHI), which contain both linear seasonal patterns and nonlinear components. This study aims to describe the characteristics of monthly GHI data in Purwakarta Regency and to implement the Hybrid Seasonal Autoregressive Integrated Moving Average–Long Short-Term Memory (SARIMA–LSTM) method to produce accurate monthly GHI forecasts using monthly average daily GHI data from January 1984 to December 2025 obtained from the NASA Prediction of Worldwide Energy Resources (POWER) database. The SARIMA model was used to capture linear and seasonal patterns, while LSTM was applied to the SARIMA residuals to model the remaining nonlinear patterns based on the Terasvirta linearity test results. The best SARIMA model obtained was SARIMA(1,0,1) (0,1,1)12. Furthermore, the Hybrid SARIMA–LSTM model produced an error which were lower than those of the SARIMA model. These results indicate that the addition of LSTM to the SARIMA residuals can improve forecasting performance by capturing nonlinear residual patterns that were not optimally modeled by the linear SARIMA model. The 2026 forecasting results show a seasonal pattern consistent with the historical characteristics of the data, with the lowest GHI value predicted in February 2026 and the highest in September 2026.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Global Horizontal Irradiance, Hybrid, LSTM, Peramalan, SARIMA, Forecasting, Global Horizontal Irradiance, Hybrid, LSTM, SARIMA
Subjects:	H Social Sciences > HA Statistics > HA30.3 Time-series analysis Q Science > QA Mathematics > QA280 Box-Jenkins forecasting Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science) T Technology > TJ Mechanical engineering and machinery > TJ810.5 Solar energy
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49201-(S1) Undergraduate Thesis
Depositing User:	Andreas Heryanto
Date Deposited:	31 Jul 2026 00:46
Last Modified:	31 Jul 2026 00:46
URI:	http://repository.its.ac.id/id/eprint/140191

Actions (login required)

View Item