Research On Wind Short-Term Prediction Based On Three Dimension Real-Time Data

Wind power has become the hot research spot among the renewable energybecause of the advantages of environmental protection, safety and inexhaustible.The wind speed is intermittent and valatile so that cause the development of thewind power become a difficult problem. The problem is mainly in the aspect ofhow to reduce the influence of the safety of the electricity system、power qualityand the balance of the power supply and demand when the large capacity windpower access to the power grid. The short-term wind speed prediction is a goodmethod to solve this problem. The wind power is proportional to the three timesof the wind speed so the output of the wind farm depends on the changing ofwind speed. The accurate short-term wind speed prediction can provide basis forthe grid dispatching and make the operation of the electricity system safety、stable and efficient.Because the wind is random, the real-time wind speed data measured by thetraditional mechanical wind meters easily have large deviation. The existing windmeters are usually measuring the horizontal wind speed. There is little researchon the three dimension anemometer considering the wind turbulence character. Ifproviding the prediction of the wind speed process with the two dimension winddata when the vertical wind speed effect much on the wind speed, the error issurely very big so that impact the prediction results. The three dimension windspeed meter was designed in this paper based on the method of ultrasonic timedifference. The system doesn’t exist mechanical consumption and the accuracy isvery high to provide the real-time three dimension wind speed data for the windspeed short-term prediction. The arrangement of the three dimension ultrasonicarray was designed based on the principle of the ultrasonic time differencemeasuring method. The high precision time to digital convert chip TDC-GP21and the STC microcontroller were the core devices in the hardware circuit andthe composition and the function of the hardware circuit were analysed in detail. The software design of the system was given in the form of flowchart. In thefinal, the measurement error and the calibration method were analysed.The support vector regression machine and the particle swarm optimizationalgorithm combined method was used to predict the three hours wind speed inorder to improve the precision of the short-term wind speed prediction. Thesupport vector regression machine and particle swarm optimization algorithmtheory were introduced. The influence of the parameters C and g in the supportvector regression machine to the prediction of the outcome was discussed. Thepredict model was construced with the LIBSVM toolbox. The comparison resultsobtained by the CV method and the PSO method were analysised and the resultsshow that the PSO-SVR method is more suitable for the short-term wind speedprediction. At last, the two dimension cup anemometer and the three dimensionwind measuring meter designed in this paper were used simultaneously to collectthe wind speed. The two methods’ data were used to predict the three hours’ windspeed with the PSO-SVR method and CV-SVR method respectively. The resultsshow that the three dimension data is better than the two dimension data and thePSO-SVR method is better than the CV-SVR method in terms of the short-termwind speed prediction.