Ultra-short-term Prediction And Error Analysis Of Wind Power Based On Chaos Theory

Among the various green energy sources that have been put into use at present,wind energy has become one of the main new energy sources to replace fossil energy by virtue of its economic advantages and easy development characteristics.Large scale wind power grid connection helps to reduce the pollution of nature and is an effective means to promote energy reform.However,due to the natural random characteristics of wind energy,its power generation has volatility,which makes it difficult for people to participate in the regulation.If such wind power is directly connected to the power grid,it will affect the power supply stability of the power grid.Therefore,it is an important means to ensure the stability of power supply to deeply study the wind power forecasting methods and improve their accuracy as much as possible.To explore better ultra-short-term prediction methods,the ultra-short-term multi-step prediction is divided into three different prediction modes: direct multi-step prediction,rolling multi-step prediction and multi sampling interval multi-step prediction,and the three prediction modes are analyzed by using the extreme learning machine prediction model with fast learning speed.Based on the above analysis,a multi-step prediction mode improvement method suitable for chaos is proposed.According to the sensitive characteristics of the initial value of chaotic time series,the applicability of the above prediction modes is improved.Finally,the weighted first-order local prediction model is used for chaos prediction to verify the prediction accuracy of different prediction modes.Based on the chaotic prediction of a single variable of wind power,the influence of the operation status of different wind turbines in the wind farm on the output power is analyzed.The grey correlation degree is used to describe the correlation degree between the operation status variables of the wind farm and the wind power sequence.The status variables are sorted according to the grey correlation degree,and the two status variables with the largest correlation degree are selected: the torque of the pitch motor and the bearing temperature of the generator,together with the wind power series,it is used as the input of the prediction model.For the chaos prediction method based on the phase space reconstruction theory,the multivariable phase space reconstruction technology is used to replace the traditional single variable reconstruction,and a weighted first-order local prediction model with multivariable input is constructed,to meet the needs of chaos prediction with the wind farm operating state variables as input.From the perspective of power load,it is found that the requirements for the accuracy of wind power prediction are different at different times of the day.For the peak load,the requirements for prediction accuracy will be greatly increased.Through the above prediction model,the prediction error analysis of the total wind power of wind farms in the three northeast provinces is carried out,and the distribution of the prediction error value is counted.On this basis,the positive error and negative error in peak load are proposed.Through the analysis of the time distribution of the difference between the positive and negative errors in the number of days,a negative error reduction method is defined to modify the predicted value through the correction coefficient,to reduce the negative error in peak load without reducing the prediction accuracy.