Forecasting Significant Wave Height using RNN-LSTM Models

Martina Maria Pushpam P., V. S. Felix Enigo

202035 citationsDOI

Abstract

Forecasting significant wave height (SWH) is necessary for many coastal and ocean engineering applications. Earlier, complex numerical models for forecasting SWH were used which was resource-intensive and computationally expensive. Currently, advanced machine learning approaches such as Deep Neural Networks are widely used in many applications that involve complex big data for predicting future events. With this idea, recurrent neural networks (RNN) with long short term memory is applied for predicting SWH at intervals of 3, 6, 12, and 24 hours a day. The presentation of the forecast model is measured based on correlation coefficient and Root Mean Square Error. It was found that different architecture of the RNN-LSTM model performs better compare to the persistence model for higher lead times.

Topics & Concepts

Recurrent neural networkComputer scienceMean squared errorArtificial intelligenceDeep learningMachine learningArtificial neural networkData modelingStatisticsMathematicsDatabaseHydrological Forecasting Using AIOcean Waves and Remote SensingEnergy Load and Power Forecasting