Forecasting Wind Speed Using A Hybrid Model Of Convolutional Neural Network And Long-Short Term Memory With Boruta Algorithm-Based Feature Selection

Abstract

With the increasing pollution and exhaustion of traditional energy resources, renewable energies are gradually becoming the topic of worldwide interest, especially wind energy. Due to its fluctuation, an accurate forecast of wind speed will contribute to the stability and reliability of the power system. In this paper, we proposed a hybrid convolutional neural network and long-short term memory (CNN-LSTM) network model combined with feature selection (FS) to predict wind speed in Switzerland. First, the important features among meteorological parameters that greatly impacted on the wind speed were founded by Boruta algorithm. Then, these features were put into CNN-LSTM model to predict. Finally, the performance of the hybrid model was compared with 5 other models, namely the single models of CNN, LSTM with or without FS and the hybrid model of CNN-LSTM without FS. The results showed that the proposed model had the highest accuracy compared with other models and could be effective in wind speed forecasting with the MAPE and RMSE of 10.01% and 1.23 km/h, respectively

Keywords

• wind speed forecasting
• hybrid model
• convolutional neural network - long-short term memory
• feature selection