可再生能源2020，Vol.38Issue(9)：1187-1191,5.
基于LSTM循环神经网络的风力发电预测
Wind power forecast based on LSTM cyclic neural network
摘要
Abstract
Large-scale wind power access to the power system will cause system frequency fluctuations. Using wind speeds at different altitudes，cosine values of wind direction, temperature，humidity，and air pressure to accurately predict wind power generation data is conducive to the development of a reasonable scheduling plan. Based on the requirements of AGC automatic power generation control, this paper selects a data collection point every 15 minutes，builds a large data set, and establishes a LSTM structure-based cyclic neural network ultra-short-term wind power generation prediction model, which is updated every 15 minutes according to the latest actual collected data. The data set implements a rolling update of the predictive network. Finally, the actual data of a certain wind field is verified. The verification results show that the algorithm has high prediction accuracy and good applicability to ultra-short-term wind power generation prediction.关键词
风力发电/LSTM循环神经网络/滚动预测/超短期风力发电预测Key words
wind power generation/ LSTM-RNN/ rolling prediction/ ultra-short-term wind power forecasting分类
能源科技引用本文复制引用
王炜,刘宏伟,陈永杰,郑楠,李政,纪项钟,于广亮,康健..基于LSTM循环神经网络的风力发电预测[J].可再生能源,2020,38(9):1187-1191,5.基金项目
国网陕西省电力公司2019年科技项目(5226JY18000G). （5226JY18000G）
