基于ES-GRU-LSTM的风电场群功率预测
    点此下载全文
引用本文:王佳钰 ,郝思鹏,李森文,王腾洲,张 伟.基于ES-GRU-LSTM的风电场群功率预测[J].计算技术与自动化,2022,(3):37-41
摘要点击次数: 380
全文下载次数: 0
作者单位
王佳钰 ,郝思鹏,李森文,王腾洲,张 伟 (南京工程学院江苏 南京 211167) 
中文摘要:风电占比的不断增加对电力系统安全稳定运行带来挑战,快速、准确的风电功率预测方法至关重要。提出了一种ES-GRU-LSTM模型对风电场群功率进行预测,通过指数平滑法(ES)处理原始数据填补缺失与异常值,提高了功率数据集的可信度和平滑性,并引入训练速度快、结构较简单的门控循环单元(GRU)对预测性能好、准确性较高的长短期记忆(LSTM)神经网络进行改进,比较ES-GRU-LSTM、GRU、LSTM的预测性能和预测时间。仿真结果表明,ES-GRU-LSTM同时改善了预测精度和预测速度。
中文关键词:长短记忆神经网络  门控循环单元  风电场群  功率预测  指数平滑法
 
Power Prediction of Wind Farm Group Based on ES-GRU-LSTM
Abstract:The increasing proportion of wind power brings challenges to the safe and stable operation of the power system. To improving the rapidity and accuracy of power prediction of wind farm group, the ES-GRU-LSTM model is proposed. First, Exponential Smoothing (ES) is adopted to process raw data and correct outliers, which brings better credibility and smoothness of the power data set. Then, the fast-training and concise Gated Recurrent Unit (GRU) is applied to the Long-Short Term Memory (LSTM) neural network with remarkable prediction performance and high accuracy. Finally, the prediction performance and prediction time of ES-GRU-LSTM is compared with those of GRU and LSTM. The simulation results show that the ES-GRU-LSTM model improves both the prediction accuracy and the prediction speed.
keywords:long-short term memory  gated recurrent unit  wind farm group  power prediction  exponential smoothing
查看全文   查看/发表评论   下载pdf阅读器