空气污染指数的前馈神经网络预测方法
|
摘要
针对石家庄空气污染指数(Air Pollution Index,简称API)的预警问题,本文提出了一种基于前馈神经网络(Feedforward Neural Network,简称FNN)的预测方法。基于三层FNN逼近任意非线性函数的特性,建立了API的预测模型;同时,针对数据变化较大的问题,采用改进学习率的梯度下降算法,对预测模型进行离线训练。选用石家庄封龙山2014年10月的空气质量数据作为研究对象,仿真实验表明,基于FNN的预测模型可以有效地对API进行预警,且预测精度可达87.94%。
For the prediction and early warning of Shijiazhuang Air Pollution Index(API), this paper proposes a method for the prediction of API via Feedforward Neural Network(FNN). Based on three-layer FNN's ability in approximating to any nonlinear functions, we establish a model of API prediction; at the same time, for the problem of the various changes of data, we adopt the gradient descent algorithm with improved learning rate for the offline training of the prediction model. We selected air quality data of Shijiazhuang Fenglong Mountain in October, 2014 as object. Simulation experiments show the model based on FNN with improved learning rate can effectively make the early warning and forecast of API, and the prediction accuracy achieves 87.94%.
For the prediction and early warning of Shijiazhuang Air Pollution Index(API), this paper proposes a method for the prediction of API via Feedforward Neural Network(FNN). Based on three-layer FNN's ability in approximating to any nonlinear functions, we establish a model of API prediction; at the same time, for the problem of the various changes of data, we adopt the gradient descent algorithm with improved learning rate for the offline training of the prediction model. We selected air quality data of Shijiazhuang Fenglong Mountain in October, 2014 as object. Simulation experiments show the model based on FNN with improved learning rate can effectively make the early warning and forecast of API, and the prediction accuracy achieves 87.94%.
引文
2 Thomas M A,Devasthale A.Sensitivity of free tropospheric carbon monoxide to atmospheric weather states and their persistency:an observational assessment over the nordic countries[J].Atmospheric Chemistry and Physics,2014,14(21):11545-11555.
3 Rajab,Jasim Mohammed,Mat Jafri M Z,Lim H S.Air surface temperature correlation with greenhouse gases by using airs data over peninsular malaysia[J].Pure and Applied Geoghysics,2014,171(8):1993-2011.
6 Feng Qi,Wu Shengjun,Du Yun.Improving neural network prediction accuracy for PM10 individual air quality index pollution levels[J].Environmental Engi-neering Science,2013,30(12):725-732.
9 Chen S,Bi lling S A.Neural network for nonlinear dynamic system modeling and identification[J].Int Journal of Control,1992,56(2):319-346.
10 Malik M I,Amir F A.The early restart algorithm[J].Neural Computation MIT,2000,(12):1303-1312.
1 孙家仁,许振成,刘煜,等.气候变化对环境空气质量影响的研究进展[J].气候与环境研究,2011,16(6):805-814.
4 王芳,程水源,李明君,等.遗传算法优化神经网络用于大气污染预报[J].北京工业大学学报,2009,35(9):1230-1234.
5 郭庆春,何振芳,李力,等.人工神经网络在API预报中的应用[J].陕西农业科学报,2011,29(4):106-108.
7 黄金杰,杨桂花,马骏驰.基于高斯的大气污染评价模型[J].计算机仿真,2011,28(2):101-104.
8 郑庆子.PM2.5对环境空气质量评价结果的影响分析[J].环境保护科学,2012,38(5):69-72.
11 陈建华.基于神经网络的非线性预测控制研究及其应用[D].华北电力大学,2007:9-56.
12 徐鹏飞,等.神经网络在时间序列预测中的应用研究[J].电子技术,2010,47.
13 鲁敏,岑红蕾,王洪坤.基于LM-BP的新疆玛纳斯灌区用水量预测[J].中国农机化学报,2014,(3):75-77.
14 中华人民共和国环境保护部环境监测司号[N].中国环境状况公报,2013.
15 Haykin S.神经网络原理[M].2版.北京:机械工业出版社,2004:165.
3 Rajab,Jasim Mohammed,Mat Jafri M Z,Lim H S.Air surface temperature correlation with greenhouse gases by using airs data over peninsular malaysia[J].Pure and Applied Geoghysics,2014,171(8):1993-2011.
6 Feng Qi,Wu Shengjun,Du Yun.Improving neural network prediction accuracy for PM10 individual air quality index pollution levels[J].Environmental Engi-neering Science,2013,30(12):725-732.
9 Chen S,Bi lling S A.Neural network for nonlinear dynamic system modeling and identification[J].Int Journal of Control,1992,56(2):319-346.
10 Malik M I,Amir F A.The early restart algorithm[J].Neural Computation MIT,2000,(12):1303-1312.
1 孙家仁,许振成,刘煜,等.气候变化对环境空气质量影响的研究进展[J].气候与环境研究,2011,16(6):805-814.
4 王芳,程水源,李明君,等.遗传算法优化神经网络用于大气污染预报[J].北京工业大学学报,2009,35(9):1230-1234.
5 郭庆春,何振芳,李力,等.人工神经网络在API预报中的应用[J].陕西农业科学报,2011,29(4):106-108.
7 黄金杰,杨桂花,马骏驰.基于高斯的大气污染评价模型[J].计算机仿真,2011,28(2):101-104.
8 郑庆子.PM2.5对环境空气质量评价结果的影响分析[J].环境保护科学,2012,38(5):69-72.
11 陈建华.基于神经网络的非线性预测控制研究及其应用[D].华北电力大学,2007:9-56.
12 徐鹏飞,等.神经网络在时间序列预测中的应用研究[J].电子技术,2010,47.
13 鲁敏,岑红蕾,王洪坤.基于LM-BP的新疆玛纳斯灌区用水量预测[J].中国农机化学报,2014,(3):75-77.
14 中华人民共和国环境保护部环境监测司号[N].中国环境状况公报,2013.
15 Haykin S.神经网络原理[M].2版.北京:机械工业出版社,2004:165.