Research On Short-Term Probabilistic Prediction Of Wind Power Taking Into Account The Spatiotemporal Characteristics Of Wind Resources

In recent years,China has vigorously developed wind power generation technology to protect the earth’s ecological environment from damage while promoting economic development.Wind energy has strong volatility and uncertainty.When large-scale wind farms are connected to the grid,it will bring many adverse effects on the operation of the main power grid.Therefore,analyzing the space-time characteristics of wind resources can improve the accuracy of wind power prediction.When the output of wind turbines fluctuates,informing the power grid dispatching in advance to prepare sufficient peak load regulation capacity can provide effective guarantee for the safe and stable operation of the power system.In order to analyze the spatiotemporal characteristics of wind power clusters,a short-term forecasting method of wind power based on spatiotemporal neural network model is proposed.The numerical weather forecast data of each wind farm in the wind power cluster is reconstructed,and a time series feature map containing rich spatiotemporal information is obtained.The time series feature map is extracted by convolutional neural network,the spatial characteristics of the data are preserved,and then the long short-term memory neural network is used for training to mine the existing time series characteristics,and then obtain the final shortterm prediction result,which also verifies the internal wind power cluster.There is a space-time relationship.In order to obtain accurate power prediction results of wind power clusters,the research starts from the mechanism of wind farms,and considers the static and dynamic characteristics of wind turbines.Fit a wind speed power curve using historical wind speed and power data and compare it to a standard wind speed power curve.Through the improved FCM clustering method,the power curves of wind farms are divided into three categories,and the curve closest to the cluster center is defined as the equivalent power curve,and then the power of the wind power cluster is predicted.In order to analyze the uncertainty of wind power output,first analyze a single wind farm,calculate the error size generated in the process of point prediction,use the self-organizing neural network（SOM）clustering to divide the wind speed into four intervals,and use nonparametric kernel density estimation to fit the error distribution in the corresponding intervals.On the basis of probability theory,the spatiotemporal dependence theory is applied to the relationship between wind speed and error,and a joint probability density distribution model of wind speed and error is established.On the basis of single-station modeling,a joint probability distribution model between different NWP characteristics and errors between wind farms at multiple locations in a wind power cluster is constructed.The importance of each factor is analyzed by grey correlation,and Copula entropy is used to obtain The weight of each feature lays the foundation for short-term probability prediction of wind power clusters.In order to fully exploit the uncertainty information existing in wind power prediction.A short-term probabilistic forecasting model of wind power based on Bootstrap sampling is proposed.The probability density distribution function corresponding to each wind speed interval is discontinuous,so the error sequence corresponding to the reconstructed cumulative probability density is used as a new data sample,and the Bootstrap method is used for sampling with replacement.Under the different confidence levels set,the size of each confidence interval is calculated separately,and the comparison and analysis are carried out with the case where only the error is considered.