A Significant Wave Height and Peak Wave Period Prediction with Transformer and LSTM Approach in Cilacap, Indonesia

Wave phenomena in the ocean can fluctuate like other weather parameters, making forecasting challenging for those who carrying out the activities at sea. Wave forecasting is very necessary to support daily marine activities such as marine transportation scheduling and daily operation offshore or in the harbor, so that these activities can be carried out with careful planning and can minimize the risk of accidents. Significant wave height (SWH) and peak wave period (Tp) predictions are essential to wave forecasting. In this research, we perform a time series wave forecasting for SWH and Tp using a relatively recent deep learning model, i.e., Transformer. As a case study, we choose a location in the southern part of Java island, Indonesia, i.e., on the Cilacap coast. We also compare the Transformer results with the well-known LSTM model. Transformer is used because it has an architecture based on encoder and decoder methods which allows this model to learn data better and also train the model faster in terms of performance. Meanwhile, LSTM is used because it's ability to remember long backward historical data. The result shows that the Transformer model performs better in terms of correlation coefficient and root mean squared error than the LSTM model for Hs. At the same time, LSTM came as a better model for Tp than the Transformer.