Improved Multi-step Method And Wind Speed To Power Conversion To Improve Wind Power Prediction Effect

Smart grid and energy internet are the main directions for current power system development.The priority access of renewable energy and more reliable and accurate prediction methods are core functions of them.Wind power as a type of renewable energy without any pollution is valued in the world.However,its volatility,intermittent and randomness etc.hinder the potential of wind energy as a quality energy source.Reliable and accurate wind power prediction is not only one of the main basic information of smart grid,global energy Internet,but also helps to eliminate the shortcomings of wind energy as non-quality energy.Accurate and reliable wind forecasting,from wind speed to accurate output power of the wind turbine are two basic tasks to improve effectiveness of wind power prediction.The two basic models of the multi-step method for forecasting future wind speed time series are direct and iterative.In this paper,the better conditions of direct and iterative are obtained through simulation,and better wind speed variation conditions respectively for direct and iterative modes are obtained.Additionally,first-order low-pass filter existing between the square of efficient wind speed and output power of the horizontal axis wind turbine is obtained by researching the influence of wind turbine mechanical inertia from wind speed to conversion of output power of the horizontal axis wind turbine.The research idea in this paper belongs to the "mechanism model + identification model" strategy of non-stationary time series prediction.The specific research works are as follows:(1)Direct and iterative prediction mode optimization in multi-step strategy of wind speed prediction.According to the basic properties of wind speed in China,the test signal set manually and the actual wind speed are used to predict and simulate the signal,then the advantages and disadvantages of direct and iterative modes can be compared.Prediction models contain Least Squares Support Vector Regression,Generalized Regression Neural Network and Random Forest methods.The conclusion that the direct mode is better when wind speed changes rapidly and iterative mode is better when wind speed changes slowly is obtained through systematic intelligent predictive simulation.(2)First-order low-pass filter relationship between the square of the efficient wind speed and the output power of the wind turbine.According to two basic mechanical movements of the horizontal axis turbine,on the time scale of wind power prediction and neglecting the various secondary influence factors,the differential equation of blade rotation is established in term of the law of rotation of rigid body in rotational motion in physics.When the wind speed does not change much,this differential equation can be approximated to a first-order low-pass filter.Using the wind tunnel experimental data of Japanese scholars and the actual data of a wind field in China,it is proved that the filter has high precision.