Short-term Power Prediction Of Wind Farm Based On BP Neural Network Optimization

Wind power is a kind of green and clean energy,with its low cost and it’s convenient for people to exploit.And it is beginning to shift from the supplement energy to strategic alternative energy.In China,about 20%of the land is of abundant wind energy resource,and it has great progress and improvement both in scale and level of development.Wind power is of huge potential development.However,because of the characteristics of wind power is randomness and intermittent,which have caused its output power unstable and also brought the new challenges to the normal and stable operation of power system.Therefore,only by managing well in wind power prediction,can the effective management of the wind farm operation,thus ensure the safety of the power system and power quality.In this background,our research is based on the content of prediction method for short-term wind power.The wind power prediction method is studied and discussed by neural method.The main aspects of the work about this paper are as follows:First of all,the methods of wind power prediction are classified.After comparing the existing methods,we decide to adopt the BP neural network to predict the wind power.On the basis of introducing the principle of BP neural,the factors influencing the output of the wind power are analyzed by wind speed,sine and cosine of the wind direction.Secondly,we select the BP neural network to forecast the wind power.The historical operation data of a wind farm is used as the resource of the training data for this model,then the typical test sample data is selected to verify the accuracy of the prediction.The results show that the BP neural network if of good predictive performance,but not stable.Finally,in order to further improve the prediction accuracy,a BP neural network prediction model optimized by artificial colony algorithm is proposed.After training with the same sample data,the same typical test sample are selected to verify the accuracy of the prediction.The results manifest that this method can reduce the prediction error of BP neural work.