Forecasting of Wind Speed by Using Deep Learning for Optimal Use of the Energy Produced by Wind Farms

Wind speed forecasting is required for predicting the power production from wind farms. Information about estimated wind power production helps in better grid management and scheduling of different types of power plants connected to the grid. This paper uses deep learning Convolutional Neural Network (CNN) for wind speed forecasting. The time series data of wind speed is separated into training, validation, and testing data and the errors in forecasting are estimated. The errors of deep learning method are also compared with that of statistical ARIMA model. It is observed that CNN makes forecasting errors comparable to that of statistical model. The wind speed dataset used in this work are of a coastal site located in Gujarat (India) and are measured by an anemometer located at 80 m height.