Wave Height Forecasting Over Ocean of Things Based on Machine Learning Techniques: An Application for Ocean Renewable Energy Generation
Kamal Upreti, Sangeeta Arora, Anupam Kumar Sharma, Adesh Kumar Pandey, Kamal Kant Sharma, Mohit Dayal
Abstract
With the evolution and integration of information and communication technologies, the marine environment is being converted into a smart ocean of things. The only way to monitor the marine environment is to access marine information through satellites, radar, etc. Recently, many researchers have focused their interest on generating power from renewable energy. Among all the available renewable resources, ocean waves are attracting the interest of researchers for power generation. Therefore, this article focuses on designing a data-driven forecasting model for marine renewable energy generation applications. This article applies a novel Gini-impurity-index-based bidirectional long short-term memory model for selecting the best ocean/marine environmental factors to forecast wave height and ultimately predict power generation using the numerical model. This article presents short- and long-term forecasting results. In the experiment, four stations each are taken for both short- and long-term forecasting. The average root-mean-square error was approximately 0.17 for long-term forecasting and approximately 0.05 for short-term forecasting.