Research On Wind Power Prediction By Data Mining On Temporal And Spatial Characteristic

Wind power is an intermittent energy and has a strong stochastic nature.With the scale-up of wind power intergration,the safe and stable operation of power is facing challenges.Wind power forecasting is an effective way to improve the penetration level of wind power and ensure the stable operation of the power grid.The traditional wind power forecasting method is improved from the perspective of data mining in this paper.It mainly includes data correction,weather type classification and spatial resource matching method.Wind power temporal and spatial characteristics which includes forecast error,time and spatial characteristics is analyzed in this paper.The analysis provides a theoretical basis for the research of this paper.Wind power data sources mainly includes numerical weather report and wind farm output power.In order to improve the data quality effectively,this paper populates the missing values in the numerical weather forecast data and eliminates the bad data in wind power output before forecasting.An online updating coefficient model which is used to correct the error is established to improve practical operating adaptability.Weather type classification facilitates the separation of different feature type data.According to the parameter which is used in classification,the method can be divided into static classification and dynamic partitioning.Base on weather type classification,different forecast model is established.The comparision between single forecast model and classification model is presented.The reason which cause the difference is analyzed.Three-dimensional meteorological parameters is used in the wind power forecasting method based on spatial resources matching.The method obtained a historical data set by calculates the similarity between the current numerical weather predictions and historical weather predictions.The regional power prediction is constructed from the measures of regional power generated during moments from a historical data set in which the numerical weather predictions were similar to the current ones.The input parameter of the current method is limited.It is not conductive to the further improve of prediction accuracy.In order to solve this probelm,a novel spatial resouces matching method which consider measureing power is presented in this paper.The experiments shows that the novel method is effective.