Research On Methods Of Prediction For Wind Speed Based On Maximum Wind Power Extraction

Energy and environmental problems have become a main global sustainable development problem in the word. Wind power has been taken more and more attention with its pollution-free and reproducibility, and keeps the fastest growing energy status in the world. Because of history and geography factors, in western region of China, there are many areas and population with no electricity. The 60 million people without electricity distribution in the western region, and the most reliable solutions to problems of long-term supply, only depending on the natural energy. Gansu hexi corridor area has rich resources and the energy density has been reached 200-300W/m2, the time for the effective wind is around 70 percent. This paper research will take full use of the region advantage to alleviate the pressure of power consumption in thi sregion even then in the state. Reducing the renewableless energy and taking an important role in promoting ecological environment aspects.With ensureing the normal working stability of wind power system at the same time, as far as possible to use the wind power and maximum wind energy tracking control is always one of the main goals of the control system. According to the characteristics of wind power and Bates theory, wind energy obtained from the wind is directly related to the speed of the generator and wind velocity. So, in the wind power control system, the accurate measurement value of the wind is vital to the energy utilization. Because of the uncertainty of air power system, the traditional method with real-time measuring easily have large deviation, and the installation of anemometer also increase the cost of equipment failure, equipment operation and complexity.Choosing appropriate maximum wind tracking control methods to improve the efficiency of the wind generator is the key factor, in view of the above problems, in this paper, based on the improved particle swarm optimize (IPSO) the RBF neural network (RBF) wind prediction and based on the least squares support vector machine (LS-SVM) speed forecasting methods are developed to improve the prediction accuracy, and then to improve the velocity of wind generators wind capture efficiency. First, a non-symmetric learning factor adjusting method introduced here is to keep the balance between the global search and the local search with the great advantages of convergence property and robustness compared with standard PSO algorithm. Use IPSO to optimizate the center ci and widthσi of the RBF neural network and to determine the accurate forecasting model of neural networks, from the results of this model is applied to predict the maximum wind speed of tracking, the average prediction error than before optimizing RBF neural network has dropped by nearly 10%. Because this algorithm much dependents on the sample data and needs a large number of samples data to train the network, so this paper puts forward a new kind of LS-SVM forecasting wind model, this model requires only two RBF kernel parameters namelyγandσi to certain and the model has simple structure, the simulation results shows that the average error is 0.08774, that indicates the method can achieve the goal, and this method has the ability of small sample learning, good generalization performance and high dimensional data processing properties, and can be used in the wind farm.The studies for the establishment of New Permanent Magnet Direct-direct Wind Turbine with full copyrights laid theoretical and practical foundation.