S, Fajar Sebastian (2025) Simulasi Gelombang Laut Menggunakan Recurrent Neural Network Long-Short Term Memory (RNN-LSTM) Pada 5 Lokasi. Other thesis, Institut Teknologi Sepuluh Nopember.
![]() |
Text
04311940000079-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (2MB) | Request a copy |
Abstract
Dilakukan simulasi gelombang laut pada 5 data dengan durasi 10 tahun di lokasi berbeda menggunakan Recurrent Neural Network Long-Short Term Memory (RNN-LSTM) untuk menghasilkan peramalan deret waktu 6 jam untuk variabel tinggi gelombang signifikan (Hs), tinggi gelombang maksimum (Hmax), periode lintas nol (Tz), dan periode puncak (Tp). Algoritma RNN LSTM dipilih karena kemampuannya untuk simulasi data runtut seperti data deret waktu. Berdasarkan hasil penelitian, 5 simulasi RNN LSTM dapat menghasilkan simulasi tinggi Hs dan Hmax yang baik (MAPE di bawah 9%), Tz yang cukup baik (MAPE di bawah 11%), namun kesulitan untuk simulasi Tp (MAPE hingga 34.06%). Kesulitan pada simulasi Tp di asosiasikan dengan kecenderungan data Tp yang memiliki standar deviasi (SD) yang tinggi lebih tinggi dari variabel lainya (SD hingga 2) dan Tp tidak korelatif dengan variabel lainya. Dapat disimpulkan RNN-LSTM dapat menghasilkan simulasi Hs, Hmax, dan Tz yang dapat diandalkan namun hanya mampu simulasi Tp dengan dengan akurasi yang cukup.
==================================================================================================================================
Sea wave simulations was conducted using 5 datasets, each with a duration of 10 years from different locations, employing the Recurrent Neural Network Long-Short Term Memory (RNN-LSTM) algorithm to perform 6-hour time series forecasting for the variables significant wave height (Hs), maximum wave height (Hmax), zero-crossing period (Tz), and peak period (Tp). The RNN-LSTM algorithm was chosen due to its capability to simulate sequential data such as time series. Based on the research findings, the 5 RNN-LSTM simulations were able to produce good forecasts for Hs and Hmax (MAPE below 9%), fairly good forecasts for Tz (MAPE below 11%), but faced difficulties in simulating Tp (MAPE up to 34.06%). The difficulty in forecasting Tp is associated with the tendency of Tp data to have a higher standard deviation (up to 2) compared to other variables, and Tp being less correlated with the other variables. It can be concluded that RNN-LSTM is capable of producing reliable forecasts for Hs, Hmax, and Tz, but can only forecast Tp with moderate accuracy.
Item Type: | Thesis (Other) |
---|---|
Uncontrolled Keywords: | RNN LSTM, Machine Learning, Simulation, Significant Wave Height, Sea Waves, Simulasi, Tinggi Gelombang Signifikan, Gelombang Laut |
Subjects: | Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. T Technology > T Technology (General) > T57.62 Simulation T Technology > TC Hydraulic engineering. Ocean engineering |
Divisions: | Faculty of Marine Technology (MARTECH) > Ocean Engineering > 38201-(S1) Undergraduate Thesis |
Depositing User: | Fajar Sebastian |
Date Deposited: | 21 Jul 2025 06:03 |
Last Modified: | 21 Jul 2025 06:03 |
URI: | http://repository.its.ac.id/id/eprint/120271 |
Actions (login required)
![]() |
View Item |