Short‐time wind speed prediction based on Legendre multi‐wavelet neural network

As one of the most widespread renewable energy sources, wind is now an important part power system. Accurate and appropriate speed forecasting has essential impact on utilisation. However, due to stochastic uncertain nature energy, more accurate necessary for its stable safer This paper proposes a Legendre multiwavelet-based neural network model non-linear prediction. It combines excellent properties multi-wavelets with self-learning capability networks, which rigorous mathematical theory support. learns input-output data pairs shares weights within divided subintervals, can greatly reduce computing costs. We explore effectiveness as activation function. Meanwhile, it successfully being applied In addition, application multi-wavelet networks in hybrid decomposition-reconstruction mode prediction problems also discussed. Numerical results real sets show that proposed able achieve optimal performance high accuracy. particular, shows multi-step prediction, illustrating superiority.