An Intelligent Data-Driven Approach for Precise Tidal Unit Modeling and Prediction

Modeling and prediction of tidal units are necessary for the optimal and practical deployment of these renewables in power systems. Unfortunately, tidal units output power depends highly on the tidal waves' characteristics which are quite random in the nature. Therefore, a very precise and powerful model is needed to make sure that this task would be achieved. This article aims to propose a novel intelligent data-driven method for modeling and prediction of tidal currents including the wave current speed and direction. To this end, a hybrid method based on auto regressive integrated moving average (ARIMA) and deep learning is proposed which can predict the tidal wave behavior based on a linear-nonlinear procedure. The proposed deep learning model makes use of two adversarial neural networks to make sure that the final model would be the most optimal one considering the limited tidal data. It is an adversarial game in which two networks try to deceive each other through a recursive training process. A novel evolutionary optimization method is also proposed to improve the adjusting parameters of the model. The experimental results show the appropriate performance of the proposed model compared to other methods in the area.