Short-term Wind Power Prediction And Its Impact On Economic Dispatching Of A Microgrid

The microgrid technology is the development direction of the electric powerindustry in the future. Clean and renewable energy including wind energy is theimportant energy source for the microgrid. However the randomness, intermittencyand uncontrollability of wind energy bring about great challenges on the energyscheduling for a microgrid system.The key method to reduce the unfavorable influence for a microgrid with windturbines due to the fluctuation of wind energy lies in the fast and accurateprediction of short-term wind power. The fast and accurate prediction targetrequires a full comparison of the characteristics of various methods duringselecting a proper short-term wind power prediction method for a microgrid. Thefollowing problem is how to improve the prediction accuracy and speed after aprediction method is established. In addition, the effect of prediction accuracy ofwind power on energy scheduling of a microgrid is also a problem due to therandomness of wind energy.Aiming at the problems mentioned above existing in a microgrid system withwind turbines, selection of prediction methods, rational configuration of modelparameters, the improvement of prediction models and the effect of predictionresults of wind power on energy scheduling of a microgrid are researched in thispaper. The main research work and innovation results are as follows:(1) The major prediction methods which are artificial neural network andsupport vector machine methods are selected out according to the consistency ofthe prediction time scale and the real-time characteristic of microgrid systems.Three representative artificial neural network methods and support vector machinemethod are chosen for detailed quantitative analysis at the angle of predictionaccuracy and speed. How the parameters configuration and sample size scale affectthe established method is also studied.(2) Different from the traditional improvement ideas, a new improvedprediction method which does not rely on the basic methods is proposed based onthe analysis of a traditional wind power prediction procedure, the predictionaccuracy is improved while the prediction speed still performs well. The proposed improved method has a good versatility because it does not depend on any basicmethods. The uncertainty prediction of wind power and the way to obtain theconfidence interval under given confidence are studied.(3) Wind power prediction values of different accuracies are given as thereference values of a microgrid system to optimize its scheduling, then thescheduling results are compared, quantitative analysis of the impact of predictionaccuracy of wind power on energy scheduling formation of a microgrid and thecorresponding unit cost of electric energy and other economic indexes are studied.