Research On Short-term Forecast Of Wind Speed And Power

With the rapid development of low carbon economy, wind power is a clean and renewable energy which is rich in resources and has the characteristics of without exploitation and transportation. It is becoming one of the most promising new energy. Wind power is now one of the renewable energy power generation technology with the fastest and the most mature development. It has the economic and technological conditions for large-scale commercial development, and also has good social and economic benefits. However, with the rapid development of wind power, wind power installed capacity is also developing rapidly, the proportion of wind power in the grid continues to increase. The inherent intermittence and fluctuation of wind energy will be a very serious threat to the quality, stability and security of the power system performance. In order to solve the impact problem to the power system when large scale wind power integrate into the power grid, effectively short-term wind power forecast has been put forward. If the wind power is forecast accurately, it will be helpful for power dispatching departments to adjust the overall scheduling timely, configure of wind turbine output reasonably and save the conventional energy power generation.This thesis uses the physical model and divides the thesis into two parts, namely the forecast of wind speed and the forecast of wind power. Wind power forecast is based on wind speed forecast. Firstly, this paper uses the WRF model to forecast the wind speed. After the analysis of the WRF framework and process, the model is used to forecast the wind speed of Ancheng wind farm in Shandong province respectively in advance of12and24hours. On this basis, this paper also compares the output results of different physical parameterization schemes in different time intervals and the results of different nesting mode. The wind speed forecast results are analyzed by error calculation. It can be concluded by the study that the output results of15minutes intervals can meet the forecast requirements. According to different prediction ageing, different combination physical schemes which are suitable for the wind field can be used to get higher accuracy. Besides, single mode with high resolution suitable for the wind field can also get better forecast results.Secondly, two methods are used to forecast the wind power--the linear method and the nonlinear method. The linear method refers to the power forecast by the establishment of the power curve, and the nonlinear method is a method using neural network for power forecast. The methods of establishing power curve include the direct method, Bean method, maximum value method, maximum likelihood method and the power function fitting method. Due to the cut-in and cut-out speed of wind power generators, different methods have different setting standards for the various speed levels. The forecast wind power can be obtained by forecast wind speed through the power curve established by these methods. After analyzing the basic principle and process of neural network, the most widely used three layers BP neural network is studied. After normalizing the original data, the BP neural network is established for short-term forecast. After that, the error of the results obtained by the methods above is calculated. This paper uses different evaluation criteria to analyze the results, including the mean absolute error, average relative error, root mean square error and the correlation coefficient of different, and finally gets more suitable method for the wind power forecast. Finally, aiming at connatural limitations of BP neural networks, the method of genetic algorithm has been put forward to optimize the weights and thresholds of the neural network. Through the example analysis, the forecast results of BP neural network model are ideal. And after the optimization of the genetic algorithm, the forecast accuracy of the model is partly improved.