Applied Research Of Support Vector Regression On Forecasting Of Short-term Wind Speed In Wind Plant

Aiming at the output fluctuation of wind farm, the research on short-term forecasting of windspeed and power have received more and more attention at home and abroad. Perfect wind speed forecast is the precondition of wind power forecast, which can adjust the scheduling, effectively mitigate the adverse effects of wind power on the grid. Based on the analysis of wind speed data in a wind farm,this paper establish Support Vector Regression (SVR) model of short-term wind speed forecast.Based on fuzzy clustering method of FCM, similar wind speed data were choosen as the training data of the model.And, analysis the parameters of SVR, Particle Swarm Optimization is adopted for parameter optimizaton,it’s validity was confirmed by comparing with the gridsearch optimization method. At last,aiming at the non-stationarity and volatility of wind speed,the wind speed forcasting method based on Empirical Mode Decomposition(EMD) and SVR is put forward. The wind speed sequence is decomposed into components of different scales by EMD,then SVR models are built up to forecast each component,and the forecasting results of each component are combined to obtain the final forcasting resut. The experiment results show that the predictive effect of this method is excellent, which can effectively reduced the adverse effects on prediction caused by the volatility of wind speed.Finally,a simple wind speed forcasting systerm is established based on VC++6.0which can realize a simple operation of visualization and man-machine communication,also can provide a preliminary reference for the further development of wind speed forcasting software.