利用时空相关性的多位置多步风速预测模型
|
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.
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.
引文
[1]叶林,赵永宁.基于空间相关性的风电功率预测研究综述[J].电力系统自动化,2014,38(14):126-135.Ye Lin,Zhao Yongning.A review on wind power prediction based on spatial correlation approach[J].Automation of Electric Power System,2014,38(14):126-135(in Chinese).
[2]陈宁,薛禹胜,丁杰,等.利用空间相关性的超短期风速预测[J].电力系统自动化,2017,41(12):124-130.Chen Ning,Xue Yusheng,Ding Jie,et al.Ultra-short term wind speed prediction using spatial correlation[J].Automation of Electric Power System,2017,41(12):124-130(in Chinese).
[3]Ak R,Fink O,Zio E.Two machine learning approaches for short-term wind speed time-series prediction[J].IEEETransactions on Neural Networks and Learning Systems,2016,27(8):1734-1747.
[4]Hu Qinghua,Su Pengyu,Yu Daren,et al.Pattern-based wind speed prediction based on generalized principal component analysis[J].IEEE Transactions on Sustainable Energy,2014,5(3):866-874.
[5]Zhang Geng,Li Hanxiong,Gan Min.Design a wind speed prediction model using probabilistic fuzzy system[J].IEEE Transactions on Industrial Informatics,2012,8(4):819-827.
[6]Mao Meiqin,Ling Jin,Chang Liuchen,et al.A novel short-term wind speed prediction based on MFEC[J].IEEE Journal of Emerging and Selected Topics in Power Electronics,2016,4(4):1206-1216.
[7]薛禹胜,陈宁,王树民,等.关于利用空间相关性预测风速的评述[J].电力系统自动化,2017,41(10):161-169.Xue Yusheng,Chen Ning,Wang Shumin,et al.Review on wind speed prediction based on spatial correlation[J].Automation of Electric Power Systems,2017,41(10):161-169(in Chinese).
[8]李莉,刘永前,杨勇平,等.基于CFD流场预计算的短期风速预测方法[J].中国电机工程学报,2013,33(7):27-32.Li Li,Liu Yongqian,Yang Yongping,et al.Short-term wind speed forecasting based on CFD pre-calculated flow fields[J].Proceedings of the CSEE,2013,33(7):27-32(in Chinese).
[9]冯双磊,王伟胜,刘纯,等.风电场功率预测物理方法研究[J].中国电机工程学报,2010,30(2):1-6.Feng Shuanglei,Wang Weisheng,Liu Chun,et al.Study on the physical approach to wind power prediction[J].Proceedings of the CSEE,2010,30(2):1-6(in Chinese).
[10]Candy B,English S J,Keogh S J.A comparison of the impact of QuikScat and WindSat wind vector products on met office analyses and forecasts[J].IEEE Transactions on Geoscience and Remote Sensing,2009,47(6):1632-1640.
[11]Focken U,Lange M,M?nnich K,et al.Short-term prediction of the aggregated power output of wind farms:a statistical analysis of the reduction of the prediction error by spatial smoothing effects[J].Journal of Wind Engineering and Industrial Aerodynamics,2002,90(3):231-246.
[12]Cellura M,Cirrincione G,Marvuglia A,et al.Wind speed spatial estimation for energy planning in Sicily:introduction and statistical analysis[J].Renewable Energy,2008,33(6):1237-1250.
[13]Hur J,Baldick R.Spatial prediction of wind farm outputs using the augmented Kriging-based model[C]//Proceedings of 2012 IEEE Power and Energy Society General Meeting.San Diego,CA,USA:IEEE,2012.
[14]Karaki S H,Chedid R B,Ramadan R.Probabilistic performance assessment of wind energy conversion systems[J].IEEE Transactions on Energy Conversion,1999,14(2):217-224.
[15]Karaki S H,Chedid R B,Ramadan R.Probabilistic performance assessment of autonomous solar-wind energy conversion systems[J].IEEE Transactions on Energy Conversion,1999,14(3):766-772.
[16]McWilliams B,Newmann M M,Sprevak D.The probability distribution of wind velocity and direction[J].Wind Engineering,1979,3(4):269-273.
[17]Bechrakis D A,Sparis P D.Correlation of wind speed between neighboring measuring stations[J].IEEETransactions on Energy Conversion,2004,19(2):400-406.
[18]Barbounis T G,Theocharis J B.Locally recurrent neural networks optimal filtering algorithms:application to wind speed prediction using spatial correlation[C]//Proceedings of IEEE International Joint Conference on Neural Networks.Montreal,Que.,Canada:IEEE,2005:2711-2716.
[19]Robert S,Foresti L,Kanevski M.Spatial prediction of monthly wind speeds in complex terrain with adaptive general regression neural networks[J].International Journal of Climatology,2013,33(7):1793-1804.
[20]Trappenberg T,Ouyang J,Back A.Input variable selection:mutual information and linear mixing measures[J].IEEE Transactions on Knowledge and Data Engineering,2006,18(1):37-46.
[21]Huang Wenhao,Song Guojie,Hong Haikun,et al.Deep architecture for traffic flow prediction:deep belief networks with multitask learning[J].IEEE Transactions on Intelligent Transportation Systems,2014,15(5):2191-2201.
[22]Zhang Pei,Wu Xiaoyu,Wang Xiaojun,et al.Short-term load forecasting based on big data technologies[J].CSEEJournal of Power and Energy Systems,2015,1(3):59-67.
[23]常亮,邓小明,周明全,等.图像理解中的卷积神经网络[J].自动化学报,2016,42(9):1300-1312.Chang Liang,Deng Xiaoming,Zhou Mingquan,et al.Convolutional neural networks in image understanding[J].Acta Automatica Sinica,2016,42(9):1300-1312(in Chinese).
[24]Chung J,Gulcehre C,Cho K H,et al.Empirical evaluation of gated recurrent neural networks on sequence modeling[J].arXiv:1412.3555,2014.
[25]Luo Xiong,Zhou Wenwen,Wang Weiping,et al.Attention-based relation extraction with bidirectional gated recurrent unit and highway network in the analysis of geological data[J].IEEE Access,2018,6:5705-5715.
[26]Zhao Rui,Wang Dongzhe,Yan Ruqiang,et al.Machine health monitoring using local feature-based gated recurrent unit networks[J].IEEE Transactions on Industrial Electronics,2018,65(2):1539-1548.
[27]Bengio Y,Simard P,Frasconi P.Learning long-term dependencies with gradient descent is difficult[J].IEEETransactions on Neural Networks,1994,5(2):157-166.
[28]Kruger N,Janssen P,Kalkan S,et al.Deep hierarchies in the primate visual cortex:what can we learn for computer vision?[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2013,35(8):1847-1871.
[29]Horn R A,Johnson C R.Topics in matrix analysis[M].New York:Cambridge University Press,1991.
[30]Ruder S.An overview of gradient descent optimization algorithms[EB/OL].(2017-7-31).https://arxiv.org/abs/1609.04747.
[31]Ripley B D.Pattern recognition and neural networks[M].Cambridge:Cambridge University Press,2007.
[32]Zhang Chunyang,Chen C L P,Gan Min,et al.Predictive deep Boltzmann machine for multiperiod wind speed forecasting[J].IEEE Transactions on Sustainable Energy,2015,6(4):1416-1425.
[33]Tieleman T,Hinton G.Lecture 6.5-RMSProp:Divide the gradient by a running average of its recent magnitude[R].Coursera:Neural Networks for Machine Learning,2012.
[34]周志华.机器学习[M].北京:清华大学出版社,2016.Zhou Zhihua.Machine learning[M].Beijing:Tsinghua University Press,2016(in Chinese).
[35]Gschwind M,Kaldewey T,Tam D K.Optimizing the efficiency of deep learning through accelerator virtualization[J].IBM Journal of Research and Development,2017,6(4-5):12:1-12:11.
[36]Melani A,Bertogna M,Bonifaci V,et al.Schedulability analysis of conditional parallel task graphs in multicore systems[J].IEEE Transactions on Computers,2017,66(2):339-353.
[2]陈宁,薛禹胜,丁杰,等.利用空间相关性的超短期风速预测[J].电力系统自动化,2017,41(12):124-130.Chen Ning,Xue Yusheng,Ding Jie,et al.Ultra-short term wind speed prediction using spatial correlation[J].Automation of Electric Power System,2017,41(12):124-130(in Chinese).
[3]Ak R,Fink O,Zio E.Two machine learning approaches for short-term wind speed time-series prediction[J].IEEETransactions on Neural Networks and Learning Systems,2016,27(8):1734-1747.
[4]Hu Qinghua,Su Pengyu,Yu Daren,et al.Pattern-based wind speed prediction based on generalized principal component analysis[J].IEEE Transactions on Sustainable Energy,2014,5(3):866-874.
[5]Zhang Geng,Li Hanxiong,Gan Min.Design a wind speed prediction model using probabilistic fuzzy system[J].IEEE Transactions on Industrial Informatics,2012,8(4):819-827.
[6]Mao Meiqin,Ling Jin,Chang Liuchen,et al.A novel short-term wind speed prediction based on MFEC[J].IEEE Journal of Emerging and Selected Topics in Power Electronics,2016,4(4):1206-1216.
[7]薛禹胜,陈宁,王树民,等.关于利用空间相关性预测风速的评述[J].电力系统自动化,2017,41(10):161-169.Xue Yusheng,Chen Ning,Wang Shumin,et al.Review on wind speed prediction based on spatial correlation[J].Automation of Electric Power Systems,2017,41(10):161-169(in Chinese).
[8]李莉,刘永前,杨勇平,等.基于CFD流场预计算的短期风速预测方法[J].中国电机工程学报,2013,33(7):27-32.Li Li,Liu Yongqian,Yang Yongping,et al.Short-term wind speed forecasting based on CFD pre-calculated flow fields[J].Proceedings of the CSEE,2013,33(7):27-32(in Chinese).
[9]冯双磊,王伟胜,刘纯,等.风电场功率预测物理方法研究[J].中国电机工程学报,2010,30(2):1-6.Feng Shuanglei,Wang Weisheng,Liu Chun,et al.Study on the physical approach to wind power prediction[J].Proceedings of the CSEE,2010,30(2):1-6(in Chinese).
[10]Candy B,English S J,Keogh S J.A comparison of the impact of QuikScat and WindSat wind vector products on met office analyses and forecasts[J].IEEE Transactions on Geoscience and Remote Sensing,2009,47(6):1632-1640.
[11]Focken U,Lange M,M?nnich K,et al.Short-term prediction of the aggregated power output of wind farms:a statistical analysis of the reduction of the prediction error by spatial smoothing effects[J].Journal of Wind Engineering and Industrial Aerodynamics,2002,90(3):231-246.
[12]Cellura M,Cirrincione G,Marvuglia A,et al.Wind speed spatial estimation for energy planning in Sicily:introduction and statistical analysis[J].Renewable Energy,2008,33(6):1237-1250.
[13]Hur J,Baldick R.Spatial prediction of wind farm outputs using the augmented Kriging-based model[C]//Proceedings of 2012 IEEE Power and Energy Society General Meeting.San Diego,CA,USA:IEEE,2012.
[14]Karaki S H,Chedid R B,Ramadan R.Probabilistic performance assessment of wind energy conversion systems[J].IEEE Transactions on Energy Conversion,1999,14(2):217-224.
[15]Karaki S H,Chedid R B,Ramadan R.Probabilistic performance assessment of autonomous solar-wind energy conversion systems[J].IEEE Transactions on Energy Conversion,1999,14(3):766-772.
[16]McWilliams B,Newmann M M,Sprevak D.The probability distribution of wind velocity and direction[J].Wind Engineering,1979,3(4):269-273.
[17]Bechrakis D A,Sparis P D.Correlation of wind speed between neighboring measuring stations[J].IEEETransactions on Energy Conversion,2004,19(2):400-406.
[18]Barbounis T G,Theocharis J B.Locally recurrent neural networks optimal filtering algorithms:application to wind speed prediction using spatial correlation[C]//Proceedings of IEEE International Joint Conference on Neural Networks.Montreal,Que.,Canada:IEEE,2005:2711-2716.
[19]Robert S,Foresti L,Kanevski M.Spatial prediction of monthly wind speeds in complex terrain with adaptive general regression neural networks[J].International Journal of Climatology,2013,33(7):1793-1804.
[20]Trappenberg T,Ouyang J,Back A.Input variable selection:mutual information and linear mixing measures[J].IEEE Transactions on Knowledge and Data Engineering,2006,18(1):37-46.
[21]Huang Wenhao,Song Guojie,Hong Haikun,et al.Deep architecture for traffic flow prediction:deep belief networks with multitask learning[J].IEEE Transactions on Intelligent Transportation Systems,2014,15(5):2191-2201.
[22]Zhang Pei,Wu Xiaoyu,Wang Xiaojun,et al.Short-term load forecasting based on big data technologies[J].CSEEJournal of Power and Energy Systems,2015,1(3):59-67.
[23]常亮,邓小明,周明全,等.图像理解中的卷积神经网络[J].自动化学报,2016,42(9):1300-1312.Chang Liang,Deng Xiaoming,Zhou Mingquan,et al.Convolutional neural networks in image understanding[J].Acta Automatica Sinica,2016,42(9):1300-1312(in Chinese).
[24]Chung J,Gulcehre C,Cho K H,et al.Empirical evaluation of gated recurrent neural networks on sequence modeling[J].arXiv:1412.3555,2014.
[25]Luo Xiong,Zhou Wenwen,Wang Weiping,et al.Attention-based relation extraction with bidirectional gated recurrent unit and highway network in the analysis of geological data[J].IEEE Access,2018,6:5705-5715.
[26]Zhao Rui,Wang Dongzhe,Yan Ruqiang,et al.Machine health monitoring using local feature-based gated recurrent unit networks[J].IEEE Transactions on Industrial Electronics,2018,65(2):1539-1548.
[27]Bengio Y,Simard P,Frasconi P.Learning long-term dependencies with gradient descent is difficult[J].IEEETransactions on Neural Networks,1994,5(2):157-166.
[28]Kruger N,Janssen P,Kalkan S,et al.Deep hierarchies in the primate visual cortex:what can we learn for computer vision?[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2013,35(8):1847-1871.
[29]Horn R A,Johnson C R.Topics in matrix analysis[M].New York:Cambridge University Press,1991.
[30]Ruder S.An overview of gradient descent optimization algorithms[EB/OL].(2017-7-31).https://arxiv.org/abs/1609.04747.
[31]Ripley B D.Pattern recognition and neural networks[M].Cambridge:Cambridge University Press,2007.
[32]Zhang Chunyang,Chen C L P,Gan Min,et al.Predictive deep Boltzmann machine for multiperiod wind speed forecasting[J].IEEE Transactions on Sustainable Energy,2015,6(4):1416-1425.
[33]Tieleman T,Hinton G.Lecture 6.5-RMSProp:Divide the gradient by a running average of its recent magnitude[R].Coursera:Neural Networks for Machine Learning,2012.
[34]周志华.机器学习[M].北京:清华大学出版社,2016.Zhou Zhihua.Machine learning[M].Beijing:Tsinghua University Press,2016(in Chinese).
[35]Gschwind M,Kaldewey T,Tam D K.Optimizing the efficiency of deep learning through accelerator virtualization[J].IBM Journal of Research and Development,2017,6(4-5):12:1-12:11.
[36]Melani A,Bertogna M,Bonifaci V,et al.Schedulability analysis of conditional parallel task graphs in multicore systems[J].IEEE Transactions on Computers,2017,66(2):339-353.