WIND ENERGY FORECSTING USING RECURRENT NEURAL NETWORKS

Abstract

Forecast of wind power is an estimation of the output production required of one or more wind turbines. Because of the variations and the probabilistic characteristics of the wind energy, forecasting it accurately becomes essential for designing reliable, economic operation and power control strategies. Changes in the nature and characteristics of the wind are probabilistic and a variety of machine learning model based on statistical parameters are used to characterize the randomness in the wind power production. The
drawbacks of different approaches include their computational complexity and their inability to adapt to timeseries processes. This paper describes Recurrent Neural Network (RNN) Long-Short Term Memory (LSTM) for time series prediction of wind power. LSTM units based RNN models have the ability to learn from the important past observations and decide whether this learned information is useful for future prediction. The experimental study showed better performance of the LSTM model compared with other traditional models.