The Establishment Of Single Models For Wind Power Prediction Based On Four Seasons

In recent years, more attention has been paid on clean and renewable energy because of the increasingly serious energy crisis, global warming and environmental pollution problems. Wind energy, as a clean and renewable energy, has been developed as a key energy around most countries on their improving energy structure and reducing environmental pollution. Wind energy is mainly utilized through wind generated power, however, wind is characterized by random, intermittent and stability, meaning that a large number of wind■power connected to the grid will bring enormous challenges to the security and stability of the power system. Therefore, accurate prediction for wind power, not only develops an appropriate scheduling strategy in advance, but also guarantees the safe and stable operation of the power system.Firstly, the basic structure and the control technology of wind turbines is overviewed and the main factors of wind power are analyzed, and the seasonal characteristics of wind speed, direction and environmental temperature are studied. Further more, a method based on quarterly modeling is proposed, and six short-term wind power prediction models are studied in this paper, which lay the foundation for the establishment of model library of wind powerprediction.The main works are organized as follows:(1) The background and significance of the wind power prediction are described; The basic structure and the control technology of wind turbines is overviewed, and the current research and common approaches of wind power prediction models are summarized.(2) The main factors of the wind power are analyzed to study its seasonal characteristics. The history data of short-term wind power are recognited by the method of data comparison, and abnormal history data are revised and repaired using Least-Squares fitting curve.(3) The theory of T-S fuzzy neural network is studied deeply. The Gauss function is replaced by the Elliptic Basis Function(EBF) as membership function to expand its receptive field, Fuzzy C-Means Clustering is used to determinethe central value of the function, and inertia is introduced to speed up the convergence of the network. The improved T-S fuzzy neural network model has been established to predict wind power for four seasons. The simulation results show that the accuracy of prediction based on improved T-S fuzzy neural network model is better than that of the whole year.(4) Further more, five models of wind power prediction are established, mainly including the general regression neural network model based on cross-validation, the Elman neural network model based on genetic algorithm, vector support machine model, multiple linear regression model, as well as improved persistence approach based on grey correlation. The method based on quarterly modeling is proposed to predict short-term wind power for four seasons. The simulation results on the MATLAB platform show that the prediction accuracy based on these models is better than that of the whole year, and the frequency of large errors are reduced, providing a strong basis to take full advantage of wind power.(5) The six single forecasting models are compared respectively and summarized the characteristics and scope of application of the models. Although the wind power prediction of each model perform an appropriate tendency, there are still large prediction error points. In order to further improve the prediction accuracy, a fusion-prediction is expected. The study of this paper lays the foundation of single model fusing of wind power prediction.