融合Savitzky-Golay滤波器的TCN-SA-BiGRU风电功率预测  

Wind power forecasting incorporating Savitzky-Golay-TCN-SA-BiGRU

作　　者：秦小晖 樊重俊[1] 付峻宇 QIN Xiaohui;FAN Chongjun;FU Junyu(Business School,University of Shanghai for Science and Technology,Shanghai 200093,China)

机构地区：[1]上海理工大学管理学院,上海200093

出　　处：《智能计算机与应用》2023年第11期166-171,共6页Intelligent Computer and Applications

基　　金：2020教育部哲学社会科学重大课题攻关项目,2020-2023(20JZD010)。

摘　　要：陆上风力发电作为主流清洁能源发电方式之一,预测其发电功率也是目前研究热点问题。本文提出融合Savitzky-Golay滤波器与基于自注意力机制的TCN-BiGRU风电功率预测模型。利用Savitzky-Golay滤波器对风电功率及相关特征数据进行降噪,随后将数据输入进由TCN时域卷积神经网络、自注意力机制模块、双向门控循环单元网络所搭建的TCN-SA-BiGRU模型中,这些模块能够更深、更快挖掘数据特征。最终预测结果显示,融合了Savitzky-Golay滤波器的模型能够有效对数据降噪,并且相较于传统单一神经网络等模型,本模型的预测性能更高。As one of the mainstream clean energy generation methods,predicting the power generated by onshore wind power is also a hot research problem at present.This paper proposes a TCN-BiGRU wind power prediction model that combines Savitzky-Golay filter and self-attentive mechanism.The Savitzky-Golay filter is used for noise reduction of wind power and related features,and the data are then fed into a TCN-SA-BiGRU model built by TCN,self-attentive mechanism,and BiGRU,which are capable of mining the data features more deeply and quickly.The final prediction results show that the model incorporating the Savitzky-Golay filter is effective in noise reduction of the data and has higher prediction performance than traditional models such as single neural networks.

关 键 词：风电功率预测 Savitzky-Golay滤波器 时域卷积神经网络(TCN) 自注意力机制 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]