Research On Prediction Of Wind Turbine Generation Based On LS-SVM

Least Squares Support Vector Machine (LS-SVM) is capable of prediction and fitting-in for the non-linear objects and wind power generation system is a nonlinear complex system. Thus, in this paper, LS-SVM is used in the research of wind turbine generation system. Model of wind turbine has been built for simulation to get characteristic curves of wind turbine and analyze factors on the impact of the wind torque. Comparing with BP neural network model, the results verify the superior performance of LS-SVM. Based on model of power coefficient, intelligent pitch-control model is offered for predicting changes of blade angle on over-rated wind speed condition. So far, it's still on the early period of theoretical research for study of short-term wind speed prediction and wind power prediction, which mainly adopt time series method and neural network method. However, in this paper, LS-SVM forecasting model is built to forecast 24 hours prediction of wind speed and wind power, using training samples and test samples of actual data. At the same time, the paper also builds particle swarm optimized-parameter model. Through comparing PSO-LSSVM model with standard model, it can be proved that PSO-LSSVM model has a better constringency than LS-SVM model, making LS-SVM model practical in actual prediction system.