Short-term Prediction Of The Wind Farm Generation Power

Energy is related to the sustainable development of the social and economic, with the appearance of the energy crisis, all the nation of the world regarded the development of renewable energy on an important position. Wind power as the fastest growing and most mature renewable energy power generation technology, there are many large-scale wind farms have access to power system. However, with the installed capacity of wind farms continuing increase, its randomness and uncertainty has become increasingly apparent for the economic, stability and power quality assurance of the power system. Wind power generation short-term prediction is one of the effective ways to solve the problem. China’s wind farms are mostly centralized, large-capacity and mostly in remote areas where the power grid construction is relatively weak, therefore, China needs short-term prediction research for wind power, but the study in this area is still in its infancy.In this article I did a more detailed study of the wind power generation short-term prediction. First, the statistical law of the data, collected from one wind farm, is studied. At the same time, make a reasonable deal and do a conversion with the data. The relationship and differences between wind turbine power generation and wind energy is also analyzed. All of these studies are a foundation for subsequent work. Then, a chaotic model, using historical power data, is build and analyzed, based on MATLAB platform. Use the RBF neural network and BP neural network model to predict. The results show that the two models have achieved good prediction. Next, respectively, based on historical wind speed data and numerical weather prediction system data to establish a time series forecasting model and least squares support vector machine model. Finally, it makes a comparative analysis of the merits and shortcomings of the three individual prediction methods. At last, on the basis of the minimum variance portfolio weights coefficient and optimal non-negative variable weighting coefficient establish combination prediction models. The results showed that combination prediction model have a very good improvement and the latter combination is more realistic, comparing with individual prediction model. The results of the study have a practical significance and application value.