Wind field forecasting using a novel method based on convolutional neural networks and bidirectional LSTM

Abstract

ABSTRACT This paper proposes a deep-learning-based wind speed forecasting model based on CNNs and BLSTM. After CNN layers there are BLSTM layers which take the high-level features from the CNN layer and capture the behaviour of data in time. Finally, the BLSTM layers are followed by a fully connected layer and loss function. The main advantage of the proposed method over previous methods is the two-level structure of the model. In this architecture, the CNN layers extract the high-level features from raw input data and feed them to the BLSTM layers which are very good at capturing the sequential pattern of the data. This feature of the network increases the accuracy and performance of the model significantly. The simulation results illustrate the accurate and reliable performance of the proposed method. Also, it is shown that the performance of the model in forecasting the U characteristics of the wind is relatively better than V characteristics.