Research On Wind Power Prediction Based On Deep Learning

Wind energy is anew type of energy which is easy to be comprehensively utilized.It is widely used in China and even the world.Wind power generation is considered to be the main form of wind energy utilization.Although wind power alleviates the problem of traditional fossil energy depletion,and is very friendly to the environment,compared with traditional thermal power generation,wind power itself has the characteristics of randomness,high instability and intermittence,which makes it difficult for wind power to connect to the stable operation of the power grid system.Forecasting the output power of wind power can effectively solve the problem of wind power access to the grid.Therefore,this paper specifically studies the high-precision wind power prediction algorithm based on the current popular deep learning algorithm.Considering that the historical data of wind power has obvious time series characteristics,this paper proposes a point prediction model of wind power at different time scales based on bidirectional long short-term memory network and attention mechanism,in which bidirectional long short-term memory network can extract the positive and negative time series characteristics of historical data The attention mechanism can make the model focus on the part of data that has a greater impact on the results.This paper collects the historical power data and meteorological data of La haute borne wind farm in northeast France.The experimental results show that the proposed method has better prediction accuracy than BP neural network,support vector machine and one-way long short-term memory network,and has good generalization ability.Due to the uncertainty of wind power output,wind power point prediction error is inevitable.Therefore,in the field of power system scheduling,it is more practical to predict the wind power output in a certain time interval.In view of this,on the basis of wind power deterministic point prediction,this paper carries out wind power interval prediction,and proposes a wind power interval prediction model based on wind power point prediction results and adaptive bandwidth kernel density estimation.Compared with point prediction,interval prediction adapts to the uncertainty characteristics of wind power output,which is of great significance for power system scheduling and other fields.