Wind Power Ultra-short-term Prediction In The Background Of Big Data

In order to cope with the energy and crisis and environmental problems,human beings are constantly striving to find new energy sources that can replace fossil fuels.Wind energy has been widely exploited and utilized for its technical maturity and realizability.But wind power has the characteristics of fluctuation and intermittence.When wind power is connected to the grid on a large scale,these characteristics of wind power will have a great impact on the safe operation of power system.Therefore,improving the accuracy of wind power prediction is of great significance to the large-scale development and utilization of wind energy and increasing the grid-connected capacity of wind power.The stochastic fluctuation of wind power is the bottleneck of improving the accuracy of wind power prediction.In view of the fluctuation of wind power,the factors causing the fluctuation of wind power are divided into external and internal factors.The external cause refers to the fluctuation of wind power caused by the change of wind speed,while the internal cause refers to the fluctuation of wind power caused by the change of wind farm state.By calculating the power loss caused by the wind farm state and the ratio of its energy to the theoretical capture power of the wind farm,it is verified that the wind farm state significantly affects the fluctuation of wind power.Therefore,the state of wind farm should be taken into account in the prediction of wind power.For the state evaluation of wind farms,it is necessary to analyze the data generated by wind turbines as a whole and linkage.Due to the complex and changeable characteristics of the environment,working conditions and state parameters in wind turbines,the parameters are interrelated and there are many types of data.Therefore,the state of the wind farm is evaluated by the random matrix theory in the large data analysis method.In view of the shortcomings of the ultra-short-term forecasting model of wind power,this paper proposes an ultra-short-term forecasting model of wind power considering the state of wind farm.In order to verify the effectiveness of wind farm state prediction,the ultra-short-term wind power prediction models are constructed by using BP neural network,RBF neural network,generalized autoregressive neural network and ELMAN neural network respectively.A distributed data platform based on Hadoop & Spark is established for wind power large data.This data platform can realize distributed storage of large wind power data.By calling MLLib module in Spark,the naive Bayesian ultra-short-term forecasting model of wind power and the random forest ultra-short-term forecasting model are realized.On the one hand,through the establishment of ultra-short-term wind power prediction models with different input information,it can be seen that multi-source data fusion can improve the accuracy of ultra-short-term wind power prediction.On the other hand,the prediction value of Spark MLLib module can provide reference for accurate ultra-short-term wind power prediction.