Short-term power prediction of a wind farm based on wavelet analysis

This paper studied the short-term prediction of wind power generating capacity by means of wavelet analysis and artificial neural network. Signal was decomposed into several sequences in different band by wavelet tranform. By building different neural network, decomposed time series were predicted separately, then the predicted results were added. Some simulations were performed using the real data from Fujin wind farm, China. The results show that the neural network model based on wavelet decomposition improves time lag problem and the mean absolute error drops from 6.99% to 6.01%, compared with the neural network model based on chaotic characteristic.