Corrected Numerical Weather Prediction-BP Neural Network Based Wind Power Short-term Prediction Research

Renewable energy has increased dramatically in the world's energy consumption structure, in which wind is considered as one of the world's most promising energy technologies. Over the past decade, with the rapid development of China's wind power industry, the consumtion. operation and grid-connection of wind power are treated as the main topics in the field of wind power research. Given that, accurate wind power prediction is not only beneficial to the grid dispatching and wind farm operation and management, but also increasing the grid's consumtion capacity of wind generation and improving the economic performance of the power system.Firstly, in the thesis, an overview of wind power prediction methods and its research status around the world were outlined. Then, combined with actual operating parameters from wind turbines in a typical wind farm, historical wind energy information and weather data records, the wind speed-wind power model was described, and the influential factors to the output were analyzed. Basing on that, the wind speed-wind power curve fitting were studied with polynomial fitting and Fourier series fitting methods.After that, wind power short-term forecasting model was established basing on the BP neural network algorithm, in which wind and weather information were taken as input elements and wind power was taken as the output of the model. A 180-days time seriers was choosen as the histotical records period for the inputs, while a 72-hours since forecasting date as the predicted period for the output.The thesis focused on studing the features of wind-speed sequence in the numerical weather prediction (NWP) information, which are the input elements of the prediction model. For different wind farms, there are some certain regularities among the overall errors from prediction wind speed and actrual records. Given that, the self-learning function of neural network can be utilized to correct the forecasted wind speed sequence, aiming to amend and improve the accuracy of predicted wind speed accuracy. Combining of the numerical weather information before and after the correction, curve fitting and BP neural network methods were separately adopted to predict the short-term wind power generation. Accoding to the wind farm condition in the case study, the results showed that when the accuracy of the predicted wind-speed from original NWP speed were low. the accuracy of wind power short-term forecast resutls using BP neural network were significantly improved after correcting the related NWP wind-speed.Finally, according to the case study, methods to achieve the prediction of the total output from a wind farm and a region are explored, and the practical work and further theoretical studies of wind power prediction are prospected.