AGgregated Wind Power Prediction For A Region

Wind power is intermittent and variable over various time scales. New problems will arise such as effects on the grid dispatch with interconnection of large capacity wind farms. How to predict wind power for real-time dispatch has become a problem for the grid. So an aggregated wind power prediction method for a region is presented. It digs the basic law of aggregated wind power with weather element as the foundation of prediction. For wind power characteristics are analyzed, it is concluded that the wind power has certain smooth effect and delay effect relative to wind speed, and regional wind power has more strong regularity. Through the analysis of the area of each wind speed and output characteristic, this paper puts forward regional effective wind speed and regional wind power concept. It analyzes the input variables of the model. Through the correlation analysis, factors having larger correlation with regional power are selected as inputs of the model and as the information characteristics of the regional wind power. The Least Squares Support Vector Machines (LS-SVM) forecasting model is established to search the historical samples with highly similar features to the forecasting power as the training samples. The fitting accuracy and generalization ability of the model depends on its relevant parameters selection. This paper adopts the ant colony algorithm to select the most optimal parameters. By analyzing a region of northeast six wind farm as an example, the region next hour wind power output is predicted. It is a straightforward approach that can improve the precision in prediction of aggregated power effectively.