Medium And Long-term Wind Power Forecasting Method Based On Resource Feature Mining

In the context of double-carbon targets and energy transformation,the construction of a new type of power system with new energy as the main body is a deepening of the connotation of the Energy Revolution,in which the interconnection of a high proportion of wind power into the grid will seriously disrupt the power system’s power supply adequacy,balance between supply and demand,etc.,the scientific medium and long-term wind power forecast can reserve the space of wind power dissipation in the level of annual and monthly power balance,and help to make the medium and long-term power generation plan,power allocation plan and equipment maintenance plan.Therefore,in order to meet the demand for the rapid development of wind power,it is necessary to break through the time scale limitation of short-term power prediction,and reserve space for wind power consumption at the medium and long-term power balance level.In this paper,considering the sufficiency of resource data,based on the existing power data of the station,the medium and long-term power prediction models are built from the perspective of resource feature mining based on the resource reanalysis data and the weather forecast results.Finally,based on the entropy weight value,the medium and long-term electricity quantity fused with multiple prediction results is obtained.First,on the basis of the reconstruction of the electricity data of coastal wind farms,the fluctuation characteristics of the medium and long-term electricity series are analyzed;then the resource characteristic parameters are constructed based on the resource reanalysis data,and the Box-Cox is used from the perspective of climate variability and error interval.The method of data transformation and differential fitting of the wind speed-power curve optimizes the process of deriving future resource characteristic parameters by the Kalman filter method;finally,the medium and longterm electricity forecasting model is built by combining the characteristic parameterelectricity conversion equation.Experimental results show that optimizing the mid-and long-term electricity prediction model from the perspective of climate variability and error interval can effectively improve the prediction accuracy of the model.Secondly,considering the important role of future climate forecast information in medium and long-term electricity forecasting,a wind energy forecasting method considering different wind energy characteristics is also presented in this paper.Taking wind energy resource weather forecast result data as input,by constructing a wind energy feature mining model,the selection of different forecast error characteristic data sets is realized,and then combined with the actual power generation data of wind farms,based on the GWO and LSTM constructs an adaptive prediction model.Validation of calculation examples shows that this method can realize wind farm and regional electricity forecasting,and the model forecasting performance is better.Finally,this paper analyzes the sequence analysis method based on resource reanalysis data using mathematical models to mine monthly resource characteristic parameters and the prediction results of the deep learning method based on weather forecast results using data-driven models to mine daily wind energy characteristic parameters,and found both The prediction errors have certain complementary characteristics,and the linear dynamic weighting method is used to fuse the prediction results of the two methods,and good experimental results have been achieved.The research in this paper can provide reference for wind power medium and long-term electricity forecasting.