Short-Term Wind Speed Prediction Based on Chaos Theory

Wind power generation features with randomness, volatility and intermittence since wind power sequence has typical chaotic characteristics. As an important method in chaos theory, phase space reconstruction can extract the dynamic intrinsic values contained in the time series. By using Takens embedding theorem, wind speed sequences are reconstructed where mutual information entropy is used to determine delay time with the benefit of reducing the correlation of each element in the phase space. A GP algorithm is used to obtain the embedding dimension afterwards. The reconstructed data is thereby used as input to predict the wind speed under the framework of machine learning methods. The measured data in a wind farm in Guizhou, China is used to test the model. The effectiveness of the proposed scheme is demonstrated by comparing with the other artificial intelligence algorithms.