Research On Wind Speed Forecasting Method Of Wind Farm Based On Support Vector Machine

Energy depletion has become a global problem,so the development and utilization of new energy remains a perennial concern.And wind energy,one of the clean,non-pollution and available new energy,has been widely used around the world.However,it will affect the safety and stability of power system and the quality of power generation because of the volatility and instability of wind power generation.The key to solve these problems is to forecast the wind speed and wind power of wind farm.The randomness of wind power can be reduced through the accurate forecast of wind speed of wind farm,and the adverse effect of wind speed variation on the power system can be mitigated effectively.Although the forecast of wind velocity has been developing rapidly in recent years,the forecast methods and forecast accuracy can be improved.Based on the establishment of several short-term wind speed forecast models,it is found that the forecast accuracy of the single wind speed forecast method is relatively insufficient.So the support vector machine and the combination forecasting model are studied and improved,and making a combined forecasting model based on least squares support vector machine(LSSVM).Firstly,the gray forecasting algorithm,the artificial neural network forecast algorithm,and the time series-Kalman filter hybrid algorithm which were selected by using the fuzzy analytic hierarchy process(AHP)in several single forecasting models,Then three single forecasting model as input and as an output value of the actual wind speed,training LSSVM,finally,obtaining the forecast function.This article also separately establish equal weight combination forecasting model and the optimal weighted combination forecasting model,and with the two combined forecasting model as a reference to analyze the combination forecasting model based on least squares support vector machine prediction performance.In this study,the prediction error of the average absolute error,the average absolute percentage error and the sum of squares of errors are used to compare the prediction performance of each model.Based on the hourly wind speed data calculated by a wind farm in Inner Mongolia as the research sample,using MATLAB to make simulation and predicted the wind speed on the basis of the models,then proving the effectiveness of this combined forecasting model.Simulation results also show that the combined forecasting model based on SVM can further improve the accuracy of wind velocity forecasting,and it also has advantages over the traditional two combined forecasting models.