Ultra-Short-Term Prediction Of Wind Power Cluster Power Considering Multi-Period Trend Perception

Hidden in the rapid development of the times is the rapid consumption of energy.How to develop and utilize new energy is the challenge of the new era.The huge amount of wind energy has the advantages of wide distribution,pollution-free and renewable,making it possible to replace primary energy in a short time.With the rapid development of wind power industry in recent years,its distribution scale has progressed to cluster and large-scale distribution.However,the intermittent,random and fluctuating characteristics of wind energy make its power output fluctuate,which affects the security and stability of the real-time operation of power system.The zoning of wind power clusters can take into account the unique space-time characteristics of wind farms in each area,which is more in line with the output characteristics of different areas.However,the characteristics of wind power make the first zoning not suitable for the full time period,so it needs to be divided into multiple time periods.In order to accurately describe the output characteristics of each region in each time period and to perceive its change trend,this paper presents an ultra-short-term prediction of wind power cluster power with multi-time trend awareness.(1)Wind power data is the basis for the follow-up study of wind power,and the existence of abnormal data affects the study of the operation law of wind turbines.High quality wind power data can provide a better environment for power prediction,which may further improve the prediction accuracy.In view of the lack of analysis on the coupling relationship between wind speed and power in traditional recognition and reconstruction algorithms,this paper presents an identification algorithm based on wind speed rise and fall characteristics and a reconstruction algorithm based on wind speed fluctuation characteristics.(2)Numeric weather prediction(NWP)contains meteorological information such as wind speed and direction at different heights in multiple regions,and it inevitably has errors with the measured data.In this paper,the distribution characteristics of NWP wind speed errors are analyzed first,and the conclusion that it is difficult to determine the error values only depending on the numerical value of wind speed is obtained.Then,the coupling relationship between the error and the numerical characteristics of NWP wind speed is analyzed.Based on this,a correction model is put forward,which uses both the numerical size of NWP wind speed and the upward and downward trend as input,to improve the correction accuracy of NWP data,and to improve the accuracy of wind power prediction.(3)To make full use of the predicted power information and numerical weather prediction information,a dynamic wind farm clustering method based on two-dimensional coordinates of power trend and wind speed fluctuation is presented.The 4-hour time scale prediction process is divided into four equal-length time scales.Balanced Iterative Reducing and Clustering Using Hierarchies(BIRCH)are used to cluster the two-dimensional coordinates of each station during each 1-hour cycle.The training set is constructed based on the result of the partition,and the power prediction of each subcluster is accomplished through the gated recurrent unit(GRU).Repeat this process until the ultra-short-term power prediction is completed for 4 hours.(4)Wind power prediction is not a single link,but a process work with multiple links interacting.At present,most of the evaluation systems only aim at the final prediction results,lacking specific analysis of the causes of errors.This paper proposes an evaluation system for the cause of wind power error.Using NWP information,measured data of meteorological information,actual operation data of wind farms,etc,the complete wind power forecasting process is deconstructed,and the impact of each link on the total forecasting error is analyzed in detail.Then the sensitivity analysis of each link and the total error is carried out to find the most effective part to improve the accuracy of wind power prediction.For subsequent prediction,the most effective accuracy improvement can be achieved by improving the link with the highest sensitivity.