The Research On Forecasting System Of Wind Speed And Wind Power

At present, with the rapid development of wind power, the wind capacity is growing rapidly, and the proportion of wind power is increasing in the power system. The prediction accuracy is of great significance to wind power due to the randomness, volatility of wind power. The methods of wind power prediction are various. However, the problem of low accuracy of prediction is ubiquitous. In this dissatation, researches were carried out about the prediction of wind power.Firstly, the main forecasting methods, the significance of the wind prediction and the development of global wind power were discussed. Then, the basic principle of three prediction methods were presented, and a group of wind data were predicted in three methods. The prediction results and error were analyzed. The sliding average method was adopted to preprocess the historical data before prediction, getting the probability interval of wind speed fluctuation. The stationary test of historical data was necessary for the time-sequence prediction method. The support vector machine and neural network should classify the historical wind speed data, dividing into training samples and predicting samples before prediction. The training samples were used to train the network, and the predicting samples were to test the prediction accuracy of the network. The three forecasting models’ results indicated that the support vector machine model are more suitable for the large amount data. The neural network prediction model of the three method was optimized by genetic algorithm. The genetic algorithm was used to optimize the number of network layer nodes, the network topology and the number of iterations. The improved prediction results showed that the prediction accuracy had been improved. This dissatation compared the three prediction methods, analyzed the applicable scope and their advantages and disadvantages. Finally, the improvement strategies were put forward with a sparse Bayesian learning model. The error level and the prediction results indicated that this model improved the prediction accuracy. As for the contagious distribution of wind farm group, the predicted output volatility and the error were small.Finally, this dissatation combined two prediction methods in MATLAB GUI, designing a practical prediction system software with good man-machine interaction, which improved the visual degree of forecasting model and made prediction convenient for staff at any time. This system provided selection of prediction methods and model input parameters. Also, it provided the comparison of different prediction results and errors.