基于门控循环单元神经网络的PM_(2.5)浓度预测
|
摘要
文章首先针对延安市市监测站单站点观测数据与PM_(2.5)的关系,从中抽取了影响PM_(2.5)较为明显的14组特征数据。依据所抽取的数据,利用LSTM深度神经网络的一种变体GRU建立了未来数小时的PM_(2.5)浓度预测模型,通过仿真实验,该模型对PM_(2.5)预测有较高的一致性,可以较好地满足日常预测业务需求。
This paper firstly analyzes the relationship between single-site observation data and PM_(2.5) of Yan'an City Monitoring Station,and extracts 14 sets of characteristic data that affect PM_(2.5).Based on the extracted data, the GRU, a variant of the LSTM deep neural network, is used to establish a PM_(2.5) concentration prediction model for the next few hours. Through simulation experiments, the model has a high consistency for PM_(2.5) prediction, and goodly meet the daily forecast business needs.
This paper firstly analyzes the relationship between single-site observation data and PM_(2.5) of Yan'an City Monitoring Station,and extracts 14 sets of characteristic data that affect PM_(2.5).Based on the extracted data, the GRU, a variant of the LSTM deep neural network, is used to establish a PM_(2.5) concentration prediction model for the next few hours. Through simulation experiments, the model has a high consistency for PM_(2.5) prediction, and goodly meet the daily forecast business needs.
引文
[1]TSENG V S,WU C W,FOURNIER-VIGER P,et al.Efficient algorithms for mining top-k high utility itemsets[J].IEEE Transactions on Knowledge and Data Engineering,2016(1):54-67.
[2]BENGIO Y,SIMARD P,FRASCONI P.Learning long term dependencies with gradient descent is difficult[J].IEEE Transactions on Neural Networks,1994(2):157-166.
[3]HOCHEREITER S,SCHMIDHUBER J.Long short-term memory[J].Neural Computation,1997(8):1735-1780.
[4]DAHL G E,ACERO A.Context-dependent pre-trained deep neural networks for large-vocabulary speech recognition[J].IEEETransactions on Audio Speech&Language Processing,2011(1):30-42.
[5]HINTON G,DENG L,YU D,et al.Deep neural networks for acoustic modeling in speech recognition:the shared views of four research groups[J].IEEE Signal Processing Magazine,2012(6):82-97.
[6]ECK D,SCHMIDHUBER J.Learning the long-term structure of the blues[J].Lecture Notes in Computer Science,2002(2415):284-289.
[7]SUTSKEVER I,VINYALS O,LE Q V.Sequence to sequence learning with neural networks[J].Computer Science,2014(4):3104-3112.
[8]范竣翔,李琦,朱亚杰,等.基于RNN的空气污染时空预报模型研究[J].测绘科学,2017(7):76-83,120.
[9]韩伟,吴艳兰,任福.基于全连接和LSTM神经网络的空气污染物预测[J].地理信息世界,2018(3):34-40.
[10]GRAVES A.Long short-term memory[M].Heidelberg:Supervised Sequence Labelling with Recurrent Neural Networks,2012.
[11]CHUNG J,GULCEHRE C,CHO K,et al.Empirical evaluation of gated recurrent neural networks on sequence modeling[C].Kuching:Conference on Neural Information Processing Systems,2014.
[2]BENGIO Y,SIMARD P,FRASCONI P.Learning long term dependencies with gradient descent is difficult[J].IEEE Transactions on Neural Networks,1994(2):157-166.
[3]HOCHEREITER S,SCHMIDHUBER J.Long short-term memory[J].Neural Computation,1997(8):1735-1780.
[4]DAHL G E,ACERO A.Context-dependent pre-trained deep neural networks for large-vocabulary speech recognition[J].IEEETransactions on Audio Speech&Language Processing,2011(1):30-42.
[5]HINTON G,DENG L,YU D,et al.Deep neural networks for acoustic modeling in speech recognition:the shared views of four research groups[J].IEEE Signal Processing Magazine,2012(6):82-97.
[6]ECK D,SCHMIDHUBER J.Learning the long-term structure of the blues[J].Lecture Notes in Computer Science,2002(2415):284-289.
[7]SUTSKEVER I,VINYALS O,LE Q V.Sequence to sequence learning with neural networks[J].Computer Science,2014(4):3104-3112.
[8]范竣翔,李琦,朱亚杰,等.基于RNN的空气污染时空预报模型研究[J].测绘科学,2017(7):76-83,120.
[9]韩伟,吴艳兰,任福.基于全连接和LSTM神经网络的空气污染物预测[J].地理信息世界,2018(3):34-40.
[10]GRAVES A.Long short-term memory[M].Heidelberg:Supervised Sequence Labelling with Recurrent Neural Networks,2012.
[11]CHUNG J,GULCEHRE C,CHO K,et al.Empirical evaluation of gated recurrent neural networks on sequence modeling[C].Kuching:Conference on Neural Information Processing Systems,2014.