Research On Short-term Wind Power Forecasting Method

With the advancement of science and technology,economic development requires a large amount of energy as a material basis.Traditional fossil energy has an imbalance between the supply and demand of energy and the economy due to its limited total amount.At the same time,the burning of fossil energy will produce a large amount of harmful substances,which will endanger the global ecological environment.In order to solve the problems of fossil energy shortage and environmental degradation,the active development of renewable energy technology has become the consensus of all countries in the world.Wind energy has been vigorously developed due to its low-carbon,economical,and green characteristics,and has gradually become the main body of renewable energy.With the vigorous development of wind power technology and the application of wind power projects,how to maximize the development of wind power resources is of vital importance to the development of the global green economy.Environmental factors have an important impact on wind power generation.Wind power is not only volatile but also random,which brings difficulties to the grid connection of wind power.In order to make better use of wind power resources,increase the utilization rate of wind power resources,reduce wind abandonment,and reduce the impact of wind power generation on the stability of the power system,research and exploration of efficient and safe wind power prediction technology is currently to solve the impact of wind power uncertainty.One of the important links.How to improve the prediction accuracy of wind power point and characterize the volatility of wind power output is the current research hotspot.Over the years,humans have made rapid progress in the field of artificial intelligence,and neural network-based wind power forecasting technology has gradually become the mainstream direction of wind power forecasting.Among them,the long short-term memory network(LSTM)has been vigorously developed in the direction of time series due to its unique structural characteristics.Based on the historical data of a wind farm in Northeast China,this paper explores the application of LSTM neural network in wind power forecasting.At the same time,the quantile regression theory is used to study the uncertainty prediction of wind power.The main work of this paper is as follows:(1)First,based on the original data,research related data preprocessing methods.Analyze and mine potential key features in historical data.At the same time,a threshold-based exception processing method is used to clean up the original data and establish an effective "clean" historical data set.(2)Secondly,a short-term wind power integrated forecasting method based on CEEMD-LSTM-Adaboost is proposed.This method uses CEEMD to reduce noise and Adaboost integrated learning features to improve the prediction ability of LSTM and improve the accuracy of short-term wind power point prediction.(3)Finally,on the basis of point prediction,the method of short-term wind power interval prediction is explored and studied.A quantile regression interval prediction method based on LASSO is proposed.This method is based on historical data and uses LASSO to filter the strong correlation factors that affect wind power prediction,eliminating human subjectivity.At the same time,combined with the quantile regression theory,the short-term wind power value under different condition quantiles is predicted,and then the wind power prediction interval is generated.Quantile interval prediction based on LASSO provides the fluctuation range of future wind power power,effectively describes the uncertainty of wind power power,and provides powerful guidance for power system dispatch.