Wind Speed Time Series Prediction Using a Single Dendritic Neuron Model

The chaotic and intrinsic complexity of wind speed time series calls for an appropriate model to accurately predict the moving tendency. In this study, we proposed a single dendritic neuron model (S-DNM) using a back-propagation algorithm to accomplish wind speed forecasting. First, based on mutual information method and false nearest neighbor method, the time delay and embedding dimension are calculated. Second, the phase space is reconstructed by time delay and embedding dimension, and the characteristics of wind speed time series are analyzed. Then, the maximum Lyapunov exponent is applied to confirm the chaotic properties of the wind speed time series. Finally, using wind speed data from Sotavento located in Galicia, Spain, the performance of the forecasting method is evaluated for short-term horizons (1 hour ahead). Experimental results show that the proposed S-DNM performed better than the traditional ELMAN model and classic multi-layered perceptron network. Thus, it is concluded that the proposed model is suitable for wind speed forecasting.