In modern times, the demand for accurate weather predictions is on the rise.
In professional windsurfing, prior knowledge on the wind’s speed and direction is used for planning the optimal route at sea. If the wind’s direction is static, the surfer can sail vertically to the wind, and then use the wind as a boost to the target. In real life, the wind shifts all the time, and so dynamic pre-planning is required to win the competition.
In this project we will attempt to provide this forecast for the Israeli Windsurfing team in the 2020 Tokyo Olympic Games, using Deep Learning. The required forecast must be in a higher time resolution and favorably in a higher accuracy than pre-existing forecasting models.
We will examine different architectures of Deep Learning Nets and train configurations. We will try to isolate interesting characteristics of the data and exploit them in our favor.
Finally, we will compare our performance to pre-existing models, and suggest future research directions.