Tryastuti, Firdha Amelia (2018) Model Hibrida ARIMAX dan Deep Neural Network untuk Peramalan Beban Listrik Jangka Pendek Area Jawa Tengah dan DIY. Undergraduate thesis, Institut Teknologi Sepuluh Nopember.
Preview |
Text
06211440000075-Undergraduate_Theses.pdf - Accepted Version Download (19MB) | Preview |
Abstract
Listrik yang dibangkitkan harus langsung digunakan oleh konsumen sehinga diperlukan perencanaan dalam bentuk peramalan beban listrik. Terdapat berbagai macam model peramalan yang telah dikembangkan untuk melakukan peramalan beban listrik. Penelitian ini bertujuan untuk meramalkan beban listrik per setengah jam di wilayah Jateng dan DIY. Model hibrida ARIMAX-DNN akan dibandingkan dengan model ARIMAX dan model Deep Neural Network. Ketiga model ini akan diterapkan pada kajian simulasi dan kajian terapan untuk data beban listrik per setengah jam wilayah Jawa Tengah dan DIY periode 1 Januari 2013 sampai 31 Desember 2017 Untuk membandingkan ketiga model tersebut akan digunakan evaluasi kebaikan model RMSEP dan sMAPEP. Hasil pada kajian simulasi menunjukkan bahwa ARIMAX merupakan model terbaik pada horizon short dan model Hibrida ARIMAX-DNN merupakan model terbaik pada horizon medium dan long. Sedangkan pada data yang mengandung noise nonlinear model Hibrida ARIMAX-DNN merupakan model terbaik pada horizon short, medium,dan long. Pada kajian terapan menunjukkan model DNN baik untuk horizon short, medium dan long.
=========================================================
Electricity generated must be directly used by consumers so that the necessary planning in the form of electricity load forecasting. There are various forecasting models that have been developed to forecast electric loads. This study aims to predict electricity load per half hour in Central Java and DIY. This study will be forecasting with ARIMAX model on the first level and Deep Neural Network at the second level. The Hybrid ARIMAX-DNN model will be compared with ARIMAX model and Deep Neural Network model. These three models will be applied to simulation studies and applied studies for electricity load data per half hour of Central Java and DIY for the period January 1, 2013 to December 31, 2017. In this study each model will be applied to 48 series of 1826 data for each series. Evaluation of the goodness of the model used to compare the three models is RMSEP and sMAPEP. The results of the simulation study show that ARIMAX is the best model on the short horizon and the ARIMAX-DNN Hybrid model is the best model on medium and long horizons. While on the data containing nonlinear noise model Hybrid ARIMAX-DNN is the best model on short, medium, and long horizons. Applied studies show the DNN model on short, medium and long horizons.
| Item Type: | Thesis (Undergraduate) |
|---|---|
| Uncontrolled Keywords: | ARIMAX, Beban Listrik, Deep Neural Network, Peramalan Jangka Pendek, RMSEP, sMAPEP |
| Subjects: | H Social Sciences > HA Statistics > HA30.3 Time-series analysis |
| Divisions: | Faculty of Mathematics, Computation, and Data Science > Statistics > 49201-(S1) Undergraduate Thesis |
| Depositing User: | Firdha Amelia Tryastuti |
| Date Deposited: | 06 Jul 2021 10:54 |
| Last Modified: | 06 Jul 2021 10:54 |
| URI: | http://repository.its.ac.id/id/eprint/57147 |
Actions (login required)
 |
View Item |