Pendekatan Percentile Error Bootstrap Pada Model Double Seasonal Holt-Winters, Double Seasonal Arima, Dan Naïve Untuk Peramalan Beban Listrik Jangka Pendek Area Jawa Timur-Bali

Khusna, Hidayatul (2015) Pendekatan Percentile Error Bootstrap Pada Model Double Seasonal Holt-Winters, Double Seasonal Arima, Dan Naïve Untuk Peramalan Beban Listrik Jangka Pendek Area Jawa Timur-Bali. Undergraduate thesis, Institut Technology Sepuluh Nopember.

[img]
Preview
Text
1311100061-Undergraduate Thesis.pdf - Published Version

Download (2MB) | Preview

Abstract

Listrik memiliki karakteristik tidak dapat disimpan. Karena itu, listrik harus dibangkitkan hanya jika diperlukan. PT. PLN perlu melakukan pengoptimalan pendistribusian listrik hingga skala operasional melalui peramalan beban listrik per setengah jam. Interval prediksi pada model double seasonal Holt-Winters (DSHW) tidak dapat dikonstruksi secara analitis. Jika digunakan untuk meramal jauh ke depan, model double seasonal ARIMA memiliki varians error yang semakin besar sehingga interval prediksi semakin lebar. Sementara model Naïve untuk data musiman memiliki varians error yang semakin besar setiap kelipatan periode musiman. Percentile error bootstrap merupakan metode nonparametrik yang dapat digunakan untuk mengkonstruksi interval prediksi. Data yang digunakan dalam penelitian ini adalah beban listrik jangka pendek area Jawa Timur-Bali dalam satuan Mega Watt (MW) untuk periode 1 Januari 2013 hingga 30 September 2013. Hasil penelitian menunjukkan bahwa model DSARIMA terbaik berdasarkan kriteria out-sample sMAPE, kriteria in-sample AIC-SBC, serta kriteria out-sample rata-rata lebar interval prediksi. Dengan demikian, dapat disimpulkan bahwa model terbaik untuk peramalan beban listrik jangka pendek area Jawa Timur-Bali adalah model DSARIMA dengan interval prediksi yang dikonstruksi menggunakan pendekatan percentile error bootstrap. ===================================================================================================== Electricity characteristic cannot be saved. Thus, electricity must be generated only if needed by customers. PT. PLN need to optimize the electricity distribution until operational scale by using half hourly electricity forecasting. The prediction interval of double seasonal Holt-Winters (DSHW) cannot be constructed analytically. Otherwise, the more step ahead to forecast, the larger variance error of DSARIMA model. Thus, the prediction interval become wider. Variance error of Naïve model for seasonal data are bigger in each seasonal period multiply. Percentile error bootstrap is one of the nonparametric methods that used to construct the prediction interval. The data that used in this research is short term electricity load demand forecasting of East Java-Bali area in Mega Watt (MW) unit for January 1st 2013 until September 30th 2014. The result shows that DSARIMA model is excellent based on out-sample sMAPE, in-sample AICSBC, and out-sample average of prediction interval width criteria. Hence, it can be concluded that the best model to forecast short term electricity load demand of East Java-Bali area is DSARIMA model which prediction interval is constructed by using percentile error bootstrap approach.

Item Type: Thesis (Undergraduate)
Additional Information: RSSt 519.535 Khu p
Uncontrolled Keywords: beban listrik, double seasonal Holt-Winters, double seasonal ARIMA, Naïve, percentile error bootstrap
Subjects: Q Science > QA Mathematics > QA276 Mathematical statistics. Time-series analysis.
Divisions: Faculty of Mathematics and Science > Statistics > (S1) Undergraduate Theses
Depositing User: Mr. Tondo Indra Nyata
Date Deposited: 28 May 2018 05:54
Last Modified: 28 May 2018 05:54
URI: http://repository.its.ac.id/id/eprint/51952

Actions (login required)

View Item View Item