摘要
兼顾时、空相关性的风速预测意义重大也极具挑战。围绕多位置、多步风速预测问题展开研究,从风速时空序列的本质出发,提出了一种"先提取空间特征,后捕捉时间依赖"的两阶段建模思路。构造了一个利用时空相关性的风速预测模型——深层时空网络(deep spatio-temporal network,DSTN)。该模型由卷积神经网络(convolutional neural network, CNN)和双向门控循环单元(bidirectional gated recurrent unit,BGRU)共同构成,并通过"端对端"的方式进行训练,具备"序列到序列"的预测能力。首先,DSTN利用底层的CNN从空间风速矩阵中提取空间特征。然后,利用BGRU捕捉来自连续时间断面的空间特征之间的时间依赖关系,进而实现对时空序列的预测。此外,还定义了针对多位置风速预测的误差指标,用以描述预测模型的总体平均性能和个体误差控制能力。以美国加利福尼亚州某风电场实测数据为算例进行分析,结果表明,DSTN能够有效利用时空相关性进行风速预测,其预测性能优于多种现有预测模型。
The wind speed prediction with spatio-temporal correlation is a task of great significance and challenges.In this paper,we are concerned with the problem of multi-step wind speed prediction for multiple sites.Based on the nature of the spatio-temporal sequences,a two-stage modeling strategy for spatio-temporal sequence prediction was proposed,i.e.,extracting the spatial features firstly and followed by capturing the temporal dependencies among these extracted spatial features.A model with deep architecture for wind speed prediction,termed the deep spatio-temporal network(DSTN),was presented.DSTN is composed of a convolutional neural network(CNN)and a bidirectional recurrent unit(BGRU),which is trained in an end-to-end(E2E)manner and capable of conducting predictions of sequence-to-sequence(S2S).Initially,the spatial features were extracted by the CNN at the bottom of DSTN.Then,the temporal dependencies among these spatial features at contiguous time sections were captured by the BGRU.Finally,the predicted wind speed was generated by the model based on the spatio-temporal correlation.Moreover,three error indices,evaluating the overall average performance and error control ability for the individual sample of the prediction model,were defined.Experiment results on real-world data from the state of California show that the proposed DSTN is capable of predicting wind speed with spatio-temporal correlation effectively,and it outperforms the prior arts.
作者
陈金富
朱乔木
石东源
李银红
ZHU Lin
段献忠
LIU Yilu
CHEN Jinfu;ZHU Qiaomu;SHI Dongyuan;LI Yinhong;ZHU Lin;DUAN Xianzhong;LIU Yilu(State Key Laboratory of Advanced Electromagnetic Engineering and Technology(School of Electrical and Electronic Engineering,Huazhong University of Science and Technology),Wuhan430074,Hubei Province,China;Department of Electrical Engineering and Computer Science,University of Tennessee,Knoxville,TN37996,U.S.A.)
出处
《中国电机工程学报》
EI
CSCD
北大核心
2019年第7期2093-2106,共14页
Proceedings of the CSEE
基金
国家重点研发计划项目(2016YFB0900101)
国家建设高水平大学公派研究生项目(201706160087)~~
关键词
深度学习
卷积神经网络
双向门控循环单元
时空相关性
多位置、多步风速预测
“端到端”学习
“序列到序列”预测
deep learning
convolutional neural network(CNN)
bidirectional gated recurrent unit(BGRU)
spatio-temporal correlation
multi-step wind speed prediction for multiple sites
end-to-end(E2E)learning
sequence-to-sequence(S2S)prediction
作者简介
陈金富(1972),男,博士,副教授,研究方向为电力系统分析计算、人工智能在电力系统分析计算中的应用以及电力系统软件技术,chenjinfu@mail.hust.edu.cn;通信作者:朱乔木(1992),男,博士研究生,研究方向为大数据和人工智能技术在电力系统中的应用,clark@hust.edu.cn;石东源(1974),男,博士,教授,研究方向为电力系统分析计算、电力系统软件技术及电力系统信息安全。
