Consideration on suitable learning method in ocean parameter prediction for stable wave power generation

Abstract

In this paper, we predict the wave height and period by applying the reservoir computing model. We used 10 types of data including wave height and period as input data, and predicted the wave height and period for the next 24 hours from the 24-hour data. We tried the method of learning and predicting all the data with the same model and the method of learning and predicting the model separately for each data group classified according to the characteristics of the data, and compared the accuracy. For the learning model, we adopted reservoir computing, which is a framework based on recurrent neural network (RNN).