Abstract

Wind power is credited with large generation capacity among the different types of renewable energy. Increased growth in wind power generation calls for accuracy in wind speed forecasting as it intermittent on various time scales. Due to its stochastic nature, many models have been developed for accurate prediction of wind speed. This paper proposes an accurate prediction model with deep learning using LSTM (Long Short-Term Memory) technique. The model is trained with real time data collected from a wind farm and it leads to a day ahead prediction of wind speed. Another model is developed with recurrent Neural Network and trained with the same data. The Mean Average Percentage Error (MAPE) for both the models are compared. By analyzing the results, the error in prediction of the wind speed is very minimal in Deep Learning using LSTM model. This is due to their nature of selectively remembering patterns for a long duration of time over the traditional neural network.