Short-term wind speed forecasting considering heteroscedasticity

Wind speed prediction of the wind farm is an important basis for the power department to make generation scheduling more scientific. It is directly related with safe, stable and economic operation of power system. Aiming at the characteristics such as non-stationary, non-linear and heteroscedasticity of wind speed sequence, historical wind speed sequence was firstly decomposed by wavelet transform and reconstructed into approximate wind speed sequence and detailed wind speed sequence respectively. Secondly, on the basis of detailed analysis of each subsequence, corresponding ARMA-ARCH comprehensive forecasting models or ARMA forecasting models are established respectively which can closely exploit and simulate the law of variation of wind speed. At last, the sub-sequences are combined on the basis of equal weight. A practical example is used to test the new improved method that Which has proved the method in this paper a effective way to improve prediction accuracy.