Peramalan Beban Konsumsi Listrik Jangka Pendek Jawa Timur dengan Functional Regression Time Varying Coefficient Model

Sinurat, Paskalis (2026) Peramalan Beban Konsumsi Listrik Jangka Pendek Jawa Timur dengan Functional Regression Time Varying Coefficient Model. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6003241028_Master_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (9MB) | Request a copy

Abstract

Beban konsumsi listrik memiliki karakteristik berfrekuensi tinggi dengan pola musiman ganda yaitu harian dan mingguan sehingga menjadi tantangan dalam peramalan beban listrik jangka pendek. Penelitian ini menerapkan Functional Regression Time-Varying Coefficient Model (FR-TVCM) untuk meramalkan beban listrik Jawa Timur dan membandingkannya dengan Hybrid Time Series Regression dan Double Seasonal ARIMA (DSARIMAX). Data yang digunakan adalah beban listrik per setengah jam dari PT PLN UP2B Jawa Timur dengan periode 1 Januari 2020 hingga 6 Januari 2025. Model FR-TVCM diestimasi dengan local linear kernel dan dibandingkan kinerja antar bandwidth dan kernel serta eksplorasi input variabel dummy. Model terbaik didapat menggunakan lag 1, 7, 8, 14, 15, kernel Gaussian, dan bandwidth 0,086 yang menghasilkan MAPE 2,69% dan sMAPE 2,73% memiliki kinerja yang sebanding dengan Hybrid DSARIMAX dengan MAPE 2,49% dan sMAPE 2,47%. Keunggulan FR-TVCM terdapat pada koefisien yang bervariasi sepanjang waktu intraharinya yang dapat diinterpretasikan. Model FR-TVCM selanjutnya digunakan untuk meramalkan beban tujuh hari ke depan pada 7 Januari 2025 hingga 13 Januari 2025.
======================================================================================================================================
Electricity consumption load is a high-frequency data with dual seasonal patterns becomes a challenge for short-term electricity load forecasting. This study applies the Functional Regression Time-Varying Coefficient Model (FR-TVCM) to forecast the electricity load in East Java and compares it with the Hybrid Time Series Regression and Double Seasonal ARIMA (DSARIMAX) models. The data used consists of half-hourly electricity demand data from PT PLN UP2B East Java with period from 1 January 2020 to 6 January 2025. The FR-TVCM model was estimated using a local linear kernel and the performance was compared across different bandwidths and kernels while also exploring the use of dummy variables. The best FRTVCM model have configuration using lags of 1, 7, 8, 14 and 15 with Gaussian kernel and a bandwidth of 0.086 yielding a MAPE of 2.69% and an sMAPE of 2.73% this performance was comparable to benchmark model Hybrid DSARIMAX which had a MAPE of 2.49% and an sMAPE of 2.47%. The advantage of the FR-TVCM lies in its coefficients, which vary over the course of the intraday period and are interpretable. The FR-TVCM model was subsequently used to short-term forecast electricity load from 7 January 2025 to 13 January 2025.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Electricity Consumption, Electricity Load, Forecasting, Time Varying Coefficient Model, Time Series
Subjects:	Q Science Q Science > QA Mathematics Q Science > QA Mathematics > QA276 Mathematical statistics. Time-series analysis. Failure time data analysis. Survival analysis (Biometry)
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49101-(S2) Master Thesis
Depositing User:	Paskalis Sinurat
Date Deposited:	04 Aug 2026 10:11
Last Modified:	04 Aug 2026 10:11
URI:	http://repository.its.ac.id/id/eprint/143664

Actions (login required)

View Item