Application Of Intelligent Neural Networks On Large Scale Wind Farm Short-Term Power Forecasting

China’s Wind Power has been gradually entering a stage of rapid development. Due to the random and intermittent nature of wind power, it brings about a great challenge to the security and stability of power system, and it causes an adverse impact on the further development of wind power, thus wind power prediction system has become an integral part of China’s wind farm development. There has not mature wind power prediction system in China. In this context, an actual wind farm in Inner Mongolia has been chosen in this paper, to study a variety of models to predict short-term power of the wind farm. The main research topics in the paper include the following aspects:Based on particle swarm optimization algorithm, establish model for each fan per month to predict the wind speed and wind power from1to6hours ahead. The results show that particle swarm optimization modeling is simple while with good prediction accuracy. Three neural networks such as GRNN, wavelet and gray have been researched in the paper. According to the deficiency of the BP neural network, adaptive simulated annealing algorithm to optimize the weights and thresholds of BP neural network initial value has been put forward. The study results show that the optimized neural network prediction accuracy has larger ascension. During the study of BP neural network wind power forecast, if training data samples are few, there is the issue of imperfect learning of the neural network. However, in the case of excessive neural network learning samples, it may bring the problem of much more complex structure of the neural network, thus the final prediction results are not very good. In accordance with sector screening training samples to train the neural network, and using neural network model and using multi-sector to forecast.To improve the robustness and accuracy of the prediction system, we have established a comprehensive model based on the principle of variable weight, the results show that the variable weight model can effectively improve the robustness and prediction accuracy of the prediction system. After we have mastered the field of wind power prediction methods and theories, the establishment of wind power prediction model to predict the wind power in Inner Mongolia, the software is not only easy to use, but also user-friendly operation, friendly interface and fully meet the power sector forecast accuracy requirements, have considerable applicability to complete the basic functions of the general short-term wind power prediction software. To achieve accurate and fast implementation of information collection, transmission, processing, analysis and storage, and make the right decisions in order to coordinate the various business sectors, in order to achieve the forecast of wind power for a time, to provide the necessary information decision support system data and decision support information...