Abstract: Along with the rapid development of wind power industry in the world, the integation of large number of wind turbines with high capacity to high voltage transmission networks is a great challenge for grid operation security and power supply quality. Wind power forecasting systems are needed for wind turbine power generation. A hybrid wind power forecasting system is proposed based on numerical weather prediction and artificial neural networks. By using historical wind speed and direction data as inputs, self-organizing maps model as the pretreatment model and radial basis function network model as the forecasting model, the forecasting system was used in a wind farm in Hebei province with the correction by historical outputs of a specific wind turbine or wind power generation field. The prediction results are acceptable, which shows the feasibility of proposed model.

Key words: wind power prediction, artificial neural network, numerical weather prediction