Aldebran, Daffa Auriz Rigel (2026) Perancangan Jaringan Syaraf Tiruan Dengan Algoritma Long Short-Term Memory Untuk Prediksi Curah Hujan Sebagai Dasar Analisis Potensi Banjir Di Jalan Teluk Kumai Barat, Surabaya. Other thesis, Institut Teknologi Sepuluh Nopember.
![[thumbnail of 5009221119-Undergraduate_Thesis.pdf]](http://repository.its.ac.id/style/images/fileicons/text.png) |
Text
5009221119-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (4MB) | Request a copy |
Abstract
Kawasan Pabean Cantian, khususnya Jalan Teluk Kumai Barat, merupakan wilayah pesisir dengan elevasi tanah rata-rata 2,05 m di atas permukaan laut. Kondisi ini menjadikan daerah tersebut rawan banjir karena sistem drainase yang ada belum mampu menampung limpasan air hujan. Penelitian ini bertujuan untuk menganalisis pengaruh curah hujan terhadap potensi banjir, membangun model prediksi curah hujan berbasis Long Short-Term Memory (LSTM) multivariat, serta mengevaluasi performa model sebagai dasar sistem prediksi curah hujan untuk menentukan potensi banjir. Data meteorologi harian periode 2020–2025 diperoleh dari BMKG Perak I, meliputi curah hujan, suhu rata-rata, lama penyinaran matahari, kecepatan angin, dan arah angin. Sementara itu, data kapasitas drainase dan ketinggian banjir dikumpulkan melalui pengukuran lapangan. Model LSTM dirancang dengan dua lapisan LSTM (64 unit) dan satu lapisan dense, dilatih menggunakan rasio data latih-uji 80:20 dengan sekuens 14 hari. Evaluasi dilakukan pada epoch 50, 100, dan 200 menggunakan metrik RMSE dan MAE. Hasil terbaik diperoleh pada epoch 200 dengan RMSE 15,2881 mm dan MAE 8,35 mm, dan juga NRMSE 11,06% dan NMAE 5,06%. Model mampu memprediksi curah hujan tiga hari ke depan dengan nilai tertinggi 1,69 mm/hari. Berdasarkan Metode Rasional, intensitas hujan prediksi menghasilkan debit limpasan maksimum 0,00876 m³/detik, masih berada di bawah kapasitas drainase eksisting. Namun, pengukuran lapangan menunjukkan curah hujan aktual dapat jauh lebih tinggi, seperti 74,4 mm/hari pada 10 Februari 2026 yang menyebabkan indikator banjir mencapai 3,24 dan genangan setinggi 31 cm di drainase bagian 2. Penelitian ini berkontribusi pada pencapaian Sustainable Development Goals (SDGs), khususnya SDG 11 (Kota dan Permukiman yang Berkelanjutan) melalui pengembangan sistem prediksi berbasis kecerdasan buatan untuk mitigasi risiko banjir perkotaan, serta SDG 13 (Penanganan Perubahan Iklim) dengan pemanfaatan data hidrometeorologi untuk memperkuat ketahanan masyarakat pesisir terhadap bencana hidrologis.
======================================================================================================================================
The Pabean Cantian area, particularly Jalan Teluk Kumai Barat, is a coastal region with an average land elevation of 2.05 m above sea level. This condition makes the area prone to flooding, as the existing drainage system is unable to accommodate stormwater runoff. This study aims to analyze the influence of rainfall on flood potential, develop a multivariate Long Short-Term Memory (LSTM) rainfall prediction model, and evaluate the model’s performance as the basis for a rainfall prediction system to determine flood risk. Daily meteorological data for the period 2020–2025 were obtained from BMKG Perak I, including rainfall, average temperature, sunshine duration, wind speed, and wind direction. Meanwhile, drainage capacity and flood height data were collected through field measurements. The LSTM model was designed with two LSTM layers (64 units) and one dense layer, trained using an 80:20 train-test ratio with a 14-day sequence. Evaluation was conducted at epochs 50, 100, and 200 using RMSE and MAE metrics. The best results were achieved at epoch 200 with RMSE of 15.2881 mm and MAE of 8.35 mm, along with NRMSE of 11.06% and NMAE of 5.06%. The model was able to predict rainfall three days ahead with the highest value of 1.69 mm/day. Based on the Rational Method, predicted rainfall intensity generated a maximum runoff discharge of 0.00876 m³/s, which remains below the capacity of the existing drainage system. However, field measurements showed that actual rainfall could be significantly higher, such as 74.4 mm/day on February 10, 2026, which caused a flood indicator of 3.24 and waterlogging up to 31 cm in drainage section 2. This research contributes to the achievement of the Sustainable Development Goals (SDGs), particularly SDG 11 (Sustainable Cities and Communities) through the development of an AI-based prediction system for urban flood risk mitigation, and SDG 13 (Climate Action) by utilizing hydrometeorological data to strengthen coastal community resilience against hydrological disasters.
Actions (login required)
 |
View Item |