SUYONO, Suyono and Syakur, Abdul and Bakhtiar, Arfan (2026) ANALISIS DETERMINISTIK DAN STOKASTIK PADA PREDIKSI BEBAN TRAFO PADA KONDISI BEBAN EKSTREM. Masters thesis, UNIVERSITAS DIPONEGORO.

	Text hal Cover-1.pdf Download (100kB)
	Text hal Cover.pdf Restricted to Repository staff only Download (837kB)
	Text Bab IPendahuluan.pdf Download (167kB)
	Text Bab IITinjauan Pustaka.pdf Download (381kB)
	Text Bab IIIMetodologi Penelitian.pdf Restricted to Repository staff only Download (363kB)
	Text Bab IVHASIL DAN PEMBAHASAN.pdf Restricted to Repository staff only Download (808kB)
	Text Bab V Penutup.pdf Restricted to Repository staff only Download (69kB)
	Text Daftar Pustaka.pdf Download (107kB)
	Text Lampiran.pdf Restricted to Repository staff only Download (727kB)

Abstract

Peramalan beban transformator merupakan aspek penting dalam perencanaan operasional dan manajemen aset sistem tenaga listrik, khususnya untuk mengantisipasi risiko pembebanan tinggi dan lonjakan beban ekstrem. Penelitian ini bertujuan untuk membandingkan kinerja model peramalan deterministik dan stokastik pada beban transformator di Gardu Induk 150 kV Pekalongan. Data yang digunakan berupa beban mingguan tiga unit transformator periode Januari 2020 hingga Desember 2024 yang bersumber dari sistem SCADA, dengan data 2020–2023 sebagai data latih dan data 2024 sebagai data uji.Model yang dibandingkan terdiri atas ETS sebagai model stokastik serta enam model deterministik, yaitu Linear, Eksponensial, Logaritmik, Polinomial Orde 2, Polinomial Orde 3, dan Polinomial Orde 4. Evaluasi dilakukan menggunakan RMSE, MAPE, dan MAE, dilanjutkan dengan Uji Friedman serta analisis Prediction Interval pada tingkat kepercayaan 95% menggunakan indikator PICP dan PINAW.Hasil penelitian menunjukkan bahwa Transformator 1 memiliki rata-rata pembebanan tertinggi sebesar 81,66%, sedangkan Transformator 3 mengalami lonjakan beban ekstrem sebesar 99,96% pada Oktober 2024. Tidak terdapat satu model yang secara konsisten unggul pada seluruh transformator, dan Uji Friedman menunjukkan bahwa perbedaan performa antar model tidak signifikan secara statistik. Temuan ini mendukung prinsip parsimoni, yaitu model sederhana tetap layak digunakan pada kondisi beban stabil, sedangkan ETS lebih reliabel dalam merepresentasikan ketidakpastian beban. Secara umum, model deret waktu univariat masih memiliki keterbatasan dalam memprediksi lonjakan beban ekstrem, sehingga pengembangan model hibrida dengan variabel eksternal diperlukan untuk meningkatkan ketahanan peramalan.
Kata Kunci : peramalan beban, transformator daya, ETS, model deterministik, Uji Friedman, prediction interval

Power transformer load forecasting is an important aspect of operational planning and asset management in electric power systems, particularly for anticipating high loading conditions and extreme load spikes. This study aims to compare the performance of deterministic and stochastic forecasting models for transformer load forecasting at the 150 kV Pekalongan Substation. The data consist of weekly load records from three power transformers obtained from the SCADA system from January 2020 to December 2024, with 2020–2023 used as training data and 2024 used as testing data.
The forecasting models compared in this study include ETS as the stochastic model and six deterministic models, namely Linear, Exponential, Logarithmic, second-order Polynomial, third-order Polynomial, and fourth-order Polynomial models. Model performance was evaluated using RMSE, MAPE, and MAE, followed by the Friedman test and Prediction Interval analysis at a 95% confidence level using PICP and PINAW.
The results show that Transformer 1 had the highest average loading level of 81.66%, while Transformer 3 experienced the most extreme load spike, reaching 99.96% in October 2024. No single model consistently outperformed the others across all transformers, and the Friedman test indicated that the performance differences among the models were not statistically significant. These findings support the principle of parsimony, indicating that simpler models remain feasible under stable load conditions, while ETS provides better reliability in representing load uncertainty. Overall, univariate time-series models still have limitations in forecasting extreme load spikes; therefore, future studies should consider hybrid models incorporating external variables to improve forecasting robustness.
Keywords : load forecasting, power transformer, ETS, deterministic model, Friedman test, prediction interval.

Actions (login required)

View Item