Foreahead Grid Prediction of Regional Wind Power Based on Spatial Reduction and CNN-LSTM
New energy power generation has received more and more attention and development around the world,among which wind energy is the most common and widely used form of new energy.Wind power prediction has an important influence on the stability,reliability and economy of the power system.In order to improve the accuracy of wind power prediction,we consider the spatial dependence of wind power stations and solve the difficulty of grid prediction in the wind power grid.First,similar wind farms are divided according to their spatial characteristics,and the convergence effect of clusters is used to simplify the number of network grids.Secondly,the meteorological characteristics with the most consistent cluster power fluctuations were extracted,and the spatial characteristics of all wind farms were retained to the maximum extent.Thirdly,the input and output data form is processed to form the network format data in the spatial sense,and is trained and predicted by convolution of long and short-term memory network prediction model.Finally,the method was applied to a large-scale wind power cluster in northeast China to verify its effectiveness.The experimental results show that the proposed method is 0.24%less RMSE and 0.05%lower compared with the ungridded MAE,which effectively improves the accuracy of day-ahead power prediction of wind power cluster.
wind power predictionconvolution long short-time memory networknetwork predictionspatial correlationcluster division
NETL
NSTL
万方数据
