Wind Power Prediction Based on Recurrent Neural Network with Long Short-Term Memory Units

Abstract

Wind power is one of the most promising renewable energy sources, it is clean, safe and inexhaustible. However, predicting wind signal has always been challenging because the time series data is nonlinear, non-stationary and chaotic. In this paper, we provide a novel predicting framework including a recurrent neural network (RNN) structure model with long-short term memory (LS TM) units and an effective forecasting map adapted to different prediction horizons. We compare our new approach with concurrent methods and show that our new method is more effective in predicting wind power.