Short-term Wind Speed Forecasting Based On Temporal And Spatial Attributes

As a clean and renewable energy,wind energy can ease the energy crisis and reduce the use of fossil fuels to protect the environment.Therefore,more and more countries attach importance to it.Wind has unstable inherent characteristics,resulting in intermittent state of power generation.When wind power is connected to the grid,it causes serious damage and difficulties to the power grid.The prediction of wind power in advance may provide the necessary basis for power grid dispatching,and then advance regulation is an effective way to solve this problem,and it is of great significance.To solve this problem,this paper uses machine learning method to predict wind speed in wind farm.Specific research contents are as follows:1)The nature of the wind speed series is analyzed: it has strong randomness and is highly correlated with the 2 order of historical data,and its stationarity is not fixed,and it is less cyclical.In addition,using a single model to predict the residual is not white noise,and there is still information in residual.In order to solve this problem,we put forward two layer model,prediction of wind speed of the first layer model to obtain the residuals,using second layer model to predict the residual sequence,the two layer prediction merge to get the final prediction results.The use of two layer model of 3 geographic location of the wind speed prediction,and compared with the single model,the experimental results show that when the prediction effect of single model is not good,the two layer model prediction performance than single model has improved.2)According to the wind speed time series prediction using single data,limited information because of the prediction problem of poor performance,the study of the physical properties with temperature and pressure,wind direction and other comprehensive at the same time,time trend,proximity and periodicity,spatial correlation of multiple wind angle,put forward three methods to select the other to get to this web site,help small site prediction fusion to achieve maximum wind speed on a site of the.Solving the problem of using a single model with high feature dimensions is difficult to train effective models because of insufficient samples.3)To use a single model to deal with high dimensional features under quasi to problem,but also difficult to weather emergencies make a good prediction for the composite model,wind speed prediction is proposed,including four parts: local predictor,global predictor,integrator and mutation predictor,it can better forecast wind speed mutation.The experimental results show that the proposed method has higher prediction accuracy than common wind speed time series prediction methods.