Research On Short-term Forecast For Wind Farm Power Based On Data-driven

In recent years,the consumption rate of traditional fossil energy is beyond our imagination.Energy shortage and environmental pollution have become hot issues in the current time.In this grim situation,countries all over the world are looking for a path of developing with a lower carbon and environmental protection,and have begun to explore renewable energy power generation technologies,among which wind power technology is the most popular.Due to the randomness,volatility and intermittent nature of wind power output,it poses a hidden danger to the stable operation of the entire power grid.Therefore,the power sector has to take measures to limit the output of wind power to avoid accidents.This not only damaged the interests of wind power companies,but also caused a great waste of wind resources.If we can master the wind power output in advance,it will be beneficial to the production and operation of wind power enterprises and the security and stability of the entire power grid.Based on this practical background,this paper takes the short-term prediction of wind power as the main research content,establishes a data-driven wind power prediction model and uses wavelet neural network as the basic research method to conduct in-depth analysis and research on wind power prediction methods.Firstly,the wind power prediction system and related prediction methods at home and abroad are introduced in detail,and the problems existing in wind power prediction are pointed out.Secondly,the impact factors of wind power output are analyzed as well.At the same time,in order to ensure the reliability and integrity of the sample data,the abnormalities,missing and power-limiting data in the sample data are identified by corresponding methods.The abnormal and missing data were repaired by means of mean method,regression method and kernel density estimation method.Because the power-limiting data has the characteristics of long duration and large amount of data,the traditional repair method can't meet the demand.Therefore,a repair method based on the data of the model machine is proposed,and a good ideal effect is achieved.Then,by analyzing the spatial correlation of wind speed,it is successfully introduced into wind power prediction model.A wind power prediction model based on spatial correlation method and wavelet neural network is proposed.Which is compared with the results of wavelet neural network prediction model,the prediction accuracy has been significantly improved.On this basis,the network weight and wavelet parameters of the wavelet neural network are determined to have a great influence on the prediction results.By comparing the optimization effect between particle swarm optimization algorithm and seeker optimization algorithm,the network weight and wavelet parameters are optimized by seeker optimization algorithm.Finally,the simulation results are carried out by MATLAB software,and the error analysis of the prediction results shows that the method adopted in this paper has higher prediction accuracy.