Research On Combined Forecasting Methods For Wind Power And Its Application

The fluctuation of wind power posing serious influences to the safe, stable and econimic operation of power system, it is the main challenge of large-scale wind power integration. Wind power forecast is one of the effective means to solve the problem. However, wind power forecast is difficult because of the complex meteorological condition’s change, single forecast model couldn’t achieve high accuracy and stability. Combination forecast can effectively use all information contained in each prediction model, which is an important means of improving the forecasting accuracy as well as diversifying risks. In the thesis, research on combined forecasting methods for wind power and its application has been undertaken and following works were conducted:1. Research on single wind power forecast modelFour single wind power forecast model were established, including GA-BP model, RBF model, SVM model and CFD pre-calculated flow fields model. A rolling samples modeling method was put forward for statistical forecast model. Results show that all of the four single wind power forecast model can achieve high accuracy, root mean square error(RMSE) of the four single model is 14.89%,14.38%,13.44%, and 13.38%; The rolling samples modeling method has higher accuracy than panel samples modeling method.2. Research on fixed weight combination forecasting model for wind powerFive fixed weight combination forecasting models were established, the weight was calculated by average method, minimum sum of the squared errors(SSE) method, entropy method, maximum grey correlation method and GA-BP method. Results show that five fixed weight combination forecasting models have problems of high fitting accuracy and low extrapolating accuracy; it is favorable to improve forecast accuracy and avoid larger forecast error if reasonable combination model was used; average method and minimum SSE method have higher forecast accuracy.3. Research on variable weight combination forecasting model based on SVM model and model screen method by comprehensive evaluation indexA comprehensive evaluation index was presented, which was used to screen models. A variable weight combination forecasting model based on SVM was established. Results show that model screen method by comprehensive evaluation index can screen models more reasonably, it is favorable to reduce forecast error; The variable weight combination forecasting model based on SVM model and model screen method by comprehensive evaluation index can achieve higher accuracy than single model, RMSE was reduced by 2.15%,1.51%,0.73%,0.68% compared to CFD pre-calculated flow fields model, GA-BP model, RBF model and SVM model, which also has better performance and practicability than fixed weight combination forecasting model.4. Development of wind power forecast system based on combination forecastA wind power forecast system based on CFD pre-calculated flow fields model, SVM model and average combination forecast model has been developed, the system update train data automatically and remotely, which can make a quick response to wind farm actual operating conditions, seasonal feature and grid demand etc.