Research On Detection And Prediction Method Of Wind Power Ramping Events

With the increasing penetration of wind power in power system,the impact of wind power ramp events in power grid,namely the large fluctuations in a short time period is becoming increasingly significant.Therefore,it is of great significance to predict the ramp events in wind energy.Compared with the research on traditional wind power forecasting,the study of ramp events is still a relatively new field.In order to better study such small probability events,the characteristics of wind power ramp events are necessary to be understood.What's more,the corresponding ramp detection methods have always been one of the obstacles in the research.Based on this,this article researched from the two aspects: detection and prediction.The main contents are as follows:In the aspect of wind power ramp events detection,in order to improve the efficiency of wind power ramp events detection,the revolving door algorithm(SDT)is used to extract the segmented trend of the original wind power data and pre-extract potential wind power ramp event sections.In view of the existence of some complicated ramp events in detection,a method of trend labeling is introduced to divide the local change trend into three types from the perspective of the power change trend: upward trend,steady trend and downward trend.Finally,the 2015 year-round data of a wind farm in Shanghai is used to identify ramp events.By sorting and analyzing the starting and ending times,ramp amount,ramp direction and corresponding weather conditions of the corresponding ramp events,the effectiveness and practicability of the climbing detection method based on the combination of the revolving door algorithm and the trend characteristics are verified.In terms of wind power ramp events prediction,considering the effect of wind speed on the prediction accuracy,it is necessary to take into account the changes in wind speed.Hence,the similar wind speed curve of the similarity degree search prediction segment considering both value similarity and shape similarity are introduced,and the wind speed and corresponding power are constituted into the final similar sample cluster.In order to obtain the optimal extreme learning machine(ELM)prediction model,the adaptive cuckoo algorithm(ACS)is introduced to optimize the input weights and thresholds of ELM.The combination of similar sample clusters and ACS-ELM wind power prediction model is proposed.Finally,the proposed ramp detection algorithm is used to realize the indirect prediction of ramp events.The experimental results show that the prediction model based on the combination of similar sample clusters and ACS-ELM can not only avoid the interference of redundant information,but also save the time required to train the model.Moreover,the prediction results are closer to the actual observation data,which indirectly ensure the effective wind power ramp event prediction.Finally,a wind power prediction system suitable for the target wind field is designed.The above algorithms is integrated into the algorithm layer module of the prediction system,and it runs stably in the actual wind field.