浙江电力2026，Vol.45Issue(4)：60-72,13.DOI:10.19585/j.zjdl.202604006
基于Informer-SDT-xLSTM协同机制的风电功率爬坡事件预测
Wind power ramping prediction based on an Informer-SDT-xLSTM collaborative mechanism
摘要
Abstract
Under extreme weather conditions,wind power output fluctuates drastically within a short period,mak-ing wind power ramping highly likely to occur.However,existing studies still exhibit limited accuracy and reliabil-ity in predicting wind power ramping under extreme weather.To address this issue,a wind power ramping prediction method based on an Informer-SDT-xLSTM collaborative mechanism is proposed.An Informer neural network model is first developed for wind power forecasting.The preliminary forecasting results are then subjected to ramping detec-tion using spinning door transformation(SDT)considering bump events,from which extrema within ramping seg-ments are extracted.Subsequently,an extended long short-term memory(xLSTM)network is employed to correct the extrema detected in the ramping segments,thereby improving prediction accuracy.The proposed method is vali-dated using data from a wind farm.Results demonstrate that the model can accurately detect wind power ramping un-der extreme weather conditions and effectively reduce prediction deviations associated with such events.关键词
风电功率预测/爬坡预测/极值修正/旋转门算法/bump事件修正/长短时记忆神经网络Key words
wind power forecasting/ramping forecast/extremum correction/SDT/bump event correction/LSTM引用本文复制引用
郑子淮,杨明,于一潇,钱建国,解慧力,李乐乐..基于Informer-SDT-xLSTM协同机制的风电功率爬坡事件预测[J].浙江电力,2026,45(4):60-72,13.基金项目
国家自然科学基金(52177095) （52177095）
国网浙江省电力有限公司科技项目(5211TZ240002) （5211TZ240002）
