Model Hybrid Multiple Seasonal untuk Peramalan Beban Listrik Jangka Pendek (Studi Kasus: Beban listrik Jangka Pendek Negara Turki)

Permata, Regita Putri (2021) Model Hybrid Multiple Seasonal untuk Peramalan Beban Listrik Jangka Pendek (Studi Kasus: Beban listrik Jangka Pendek Negara Turki). Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 06211950010011_Master_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2023. Download (6MB) | Request a copy

Abstract

Peramalan beban listrik jangka pendek merupakan bagian paling penting dalam menyiapkan kebutuhan energi listrik setiap hari. Pola beban listrik jangka pendek adalah musiman harmonik (harian, mingguan, dan tahunan) serta adanya penurunan beban listrik pada hari Raya Idul Fitri dan Hari Raya Idul Adha, serta beberapa pola tidak teratur seperti musim (fall, summer, winter, spring), dan Efek pandemi
COVID-19. Dalam penelitian ini bertujuan untuk mengembangkan metode statistika dan hibrida dengan adanya variasi kalender hijriyah dan pola tidak teratur berdasarkan dua scenario, yang mana skenario 1 data out of sample mengandung
variasi kalender Hari Raya Idul Adha dan harmonik musiman. Skenario 2 memuat data out of sample hanya mengandung harmonik musiman. Data yang digunakan adalah data beban listrik perjam Negara Turki yang diakses melalui kaggle.com.
Metode yang digunakan dalam peramalan beban listrik jangka pendek adalah Dynamic Harmonic Regression serta metode hybrid yang mengakomodasi variasi kalender dan pola ireguler dengan model regresi time series. Selain itu menggunakan multiple seasonal ARIMA, pemodelan hybrid regresi time series efek variasi kalender dengan multiple seasonal ARIMA dan hybrid Neural Network sebagai pembanding. Berdasarkan hasil empiris, hasil model terbaik dari dua skenario berbeda menurut akurasi error. Skenario 1 error minimum adalah model
hybrid TSR-MSARIMA, sedangkan skenario 2 error minimum adalah hybrid TSRDHR. Model hybrid TSR-DHR memiliki kinerja yang lebih baik, dan lebih konsisten dalam meramalkan setiap periode. Dapat disimpulkan bahwa metode hybrid dapat menghasilkan hasil yang lebih akurat daripada individual peramalan pada dua skenario yang dirancang dalam penelitian ini. Evaluasi kesalahan peramalan menunjukkan bahwa semakin panjang periode out of sample, semakin besar kesalahan yang akan dihasilkan.
=====================================================================================================
Forecasting short-term electrical loads are the most important part of preparing daily electrical energy needs. Short-term electrical load patterns are seasonal harmonics (daily, weekly, and annual) as well as a decrease in electrical load on certain celebration days, namely Eid al-Fitr and Eid al-Adha, as well as several irregular patterns such as seasons (fall, summer, winter, spring), and the effects of the COVID-19 pandemic. This study aims to develop statistical and hybrid methods with variations in the hijri calendar and irregular patterns based on two scenarios,
in which scenario 1 data out of sample contains variations in the Eid al-Adha calendar and seasonal harmonics. Scenario 2 contains out of sample data only containing seasonal harmonics. The data used is Turkey's hourly electricity load
data which is accessed through kaggle.com. The method used in short-term electrical load forecasting is Dynamic Harmonic Regression and a hybrid method that accommodates calendar variations and irregular patterns with a time series regression model. In addition, it uses multiple seasonal ARIMA, time series hybrid regression modeling the effects of calendar variations with multiple seasonal
ARIMA and hybrid Neural Network as comparisons. Based on empirical results, the best model results from the two scenarios differ according to the error accuracy.
The minimum 1 error scenario is the TSR-MSARIMA hybrid model, while the 2 minimum error scenario is the TSR-DHR hybrid. The TSR-DHR hybrid model has better performance, and is more consistent in forecasting each period. It can be
concluded that the hybrid method can produce more accurate results than individual forecasting in the two scenarios designed in this study. Evaluation of forecasting
errors shows that the longer the out-of-sample period, the greater the error that will be generated

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	dynamic harmonic regression, hybrid, multiple seasonal ARIMA, peramalan jangka pendek, variasi kalender, calendar variations, short-term forecasting
Subjects:	H Social Sciences > HA Statistics > HA30.3 Time-series analysis Q Science > QA Mathematics > QA280 Box-Jenkins forecasting Q Science > QA Mathematics > QA404 Fourier series
Divisions:	Faculty of Mathematics and Science > Statistics > 49101-(S2) Master Thesis
Depositing User:	Regita Putri Permata
Date Deposited:	09 Sep 2021 08:56
Last Modified:	09 Sep 2021 08:56
URI:	http://repository.its.ac.id/id/eprint/91904

Actions (login required)

View Item