Nafik, FahmiKhuluqin (2014) Peramalan Inflow Debit Air Waduk Menggunakan Backpropagation Neural Network Dengan Input Variabel Lag (Study Kasus : Waduk Sengguruh – Kabupaten Malang). Sarjana thesis, Universitas Brawijaya.
Abstract
Cuaca yang tidak merata di Indonesia sepanjang tahun menyebabkan persediaan air yang berlebihan dimusim penghujan dan kekukarangan air dimusim kemarau. Oleh sebab itu, perlu dipertimbangkan keseimbangan debit air yang masuk (inflow) dan debit air keluar (outflow) pada waduk. Dalam penelitian ini, digunakan metode Backpropagation Neural Network untuk melakukan peramalan debit inflow menggunakan 3 data ramalan yang berbeda, yaitu harian umum, mingguan dan klasifikasi berdasarkan hari (masing-masing hari). Data input yang digunakan sebagai masukan JST Backpropagation adalah variabel lag yang diperoleh dari plot autokorelasi parsial (PACF). Berdasarkan uji coba, variabel lag yang beda nyata (keluar batas) hasil plot PACF adalah lag-1, lag-2 dan lag-3 sehingga banyak neuron pada lapisan input adalah 3 neuron. Sedangkan tahap uji coba JST Backpropagation, peramalan harian umum arsitektur (3-15-1) memiliki MSE minimum 0.00011960. Peramalan mingguan arsitektur terbaik (3-5-1) menghasilkan MSE testing 0.00007590. MAPE yang dihasilkan dari 2 arsitektur ini tergolong rendah dengan persentase 5.1431% dan 4.3371%. Sedangkan peramalan berdasarkan klasifikasi hari, peramalan hari selasa memiliki nilai MSE minimum 0.00003964 dan MAPE 3.1340 % dengan arsitektur (3-5-1). Secara umum hasil peramalan menggunakan 3 ramalan berbeda tidaklah terlalu jauh tingkat akurasinya.
English Abstract
The different weather in Indonesia in a year causes excessive water supply in rainy seasson and water less in dry seasson. Therefore, we need to considere the balance of inflow and outflow in a reservoir. This research, uses Backpropagation Neural Network method to overcast inflow using 3 different data, they are daily, weekly and each days. The data input used as Backpropagation Neural Network is lag variabel got from Partial Autocorrelation Function (PACF). Based on the trials, the lag variable of PACF plot result which were different significantly (out of border) were lag-1, lag-2 and lag-3. And as a result, it caused the number of neuron on the input layer was three neurons. On the other hand, the trial of JST Backpropagation, daily general forecasting of the architecture (3-5-1) had minimum MSE 0.00011960. The weekly forecasting of the best architecture (3-5-1) produce MSE testing 0.00007590. the MAPE produce by these two architectures was low with the percantage of 5.1431% and 4.3371%. whereas the forecasting that was based on the daily classification, the forecasting on Tuesday produced MSE minimum value 0.00003964 and MAPE 3.1340% using the architecture (3-5-1). In general, the accuracy level produced by using three different type of forecasting was not significantly different.
| Item Type: | Thesis (Sarjana) |
|---|---|
| Identification Number: | SKR/MIPA/2014/272/051405151 |
| Subjects: | 500 Natural sciences and mathematics > 510 Mathematics |
| Divisions: | Fakultas Matematika dan Ilmu Pengetahuan Alam > Matematika |
| Depositing User: | Budi Wahyono Wahyono |
| Date Deposited: | 01 Sep 2014 11:11 |
| Last Modified: | 21 Oct 2021 04:41 |
| URI: | http://repository.ub.ac.id/id/eprint/153902 |
Preview |
Text
5-BAB_III.pdf Download (1MB) | Preview |
Preview |
Text
2-AWALAN.pdf Download (1MB) | Preview |
Preview |
Text
3-BAB_I.pdf Download (1MB) | Preview |
Preview |
Text
6-BAB_IV.pdf Download (2MB) | Preview |
Preview |
Text
4-BAB_II.pdf Download (2MB) | Preview |
Preview |
Text
7-BAB_V.pdf Download (1MB) | Preview |
Preview |
Text
9-LAMPIRAN.pdf Download (3MB) | Preview |
Preview |
Text
8-DAFTAR_PUSTAKA.pdf Download (1MB) | Preview |
Preview |
Text
1-COVER_UTAMA.pdf Download (1MB) | Preview |
Actions (login required)
 |
View Item |