利用Keras构建神经网络在空气质量预测中的应用
|
摘要
利用python语言搭建了一整套空气质量神经网络预测系统,底层利用Keras设计并建立了基于tensorflow的神经网络模型。选择日平均气压、日平均气温、日平均相对湿度、日降水量、日平均风速、前1日空气质量因子监测数据等因素作为模型输入变量,分别针对广东省所有监测站点和地市的空气质量因子(PM_(2.5)、PM_(10)、NO_2、SO_2、CO、O_3、AQI)进行预测,结果表明,7个因子的地市平均相对误差值为16.15%～27.7%,地市相关系数为0.36～0.77,该模型在城市空气质量预测中具有良好的效果。
In this paper,a set of neural network system has been built for air quality prediction by python language. The bottom layer adopted Keras design and the neural network model was established based on tensorflow. The system has been applied to forecast air quality( PM2. 5,PM10,NO2,SO2,CO,O3,AQI) in all monitoring stations and prefecture-level cities of Guangdong Province,on the basis of taking the meteorological data such as daily mean pressure,daily mean temperature,daily mean relative humidity,daily precipitation,daily mean wind speed and previous day's monitoring data of air quality as the input data. The results showed that the average mean-relative-error values of prefecture-level cities were between 16. 15% and 27. 7%,and the average correlation coefficient values of prefectural-level cities were between 36. 41% and 76. 65%. These results proved that the model is highly accessible and feasible in urban air quality prediction.
In this paper,a set of neural network system has been built for air quality prediction by python language. The bottom layer adopted Keras design and the neural network model was established based on tensorflow. The system has been applied to forecast air quality( PM2. 5,PM10,NO2,SO2,CO,O3,AQI) in all monitoring stations and prefecture-level cities of Guangdong Province,on the basis of taking the meteorological data such as daily mean pressure,daily mean temperature,daily mean relative humidity,daily precipitation,daily mean wind speed and previous day's monitoring data of air quality as the input data. The results showed that the average mean-relative-error values of prefecture-level cities were between 16. 15% and 27. 7%,and the average correlation coefficient values of prefectural-level cities were between 36. 41% and 76. 65%. These results proved that the model is highly accessible and feasible in urban air quality prediction.
引文
[1]殷骏,潘梦玲.空气污染已成人类健康第一杀手[J].生态经济,2016,32(1):6-9.
[2]王晓彦,刘冰,李健军,等.区域环境空气质量预报的一般方法和基本原则[J].中国环境监测,2015,31(1):134-138.
[3]吴兑,邓雪娇,林爱兰,等.广东省空气质量预报系统[J].气象科技,2003(6):351-355.
[4]李小飞,张明军,王圣杰,等.中国空气污染指数变化特征及影响因素分析[J].环境科学,2012(6):1936-1943.
[5]朱玉强.几种空气质量预报方法的预报效果对比分析[J].气象,2004(10):30-33.
[6]陈彬彬,林长城,杨凯,等.基于CMAQ模式产品的福州市空气质量预报系统[J].中国环境科学,2012,32(10):1744-1752.
[7]任万辉,苏枞枞,赵宏德.城市环境空气污染预报研究进展[J].环境保护科学,2010(3):9-11.
[8]沈路路,王聿绚,段雷.神经网络模型在O3浓度预测中的应用[J].环境科学,2011,32(8):2231-2235.
[9]赵宏,刘爱霞,王恺,等基于GA-ANN改进的空气质量预测模型[J].环境科学研究,2009(11):1276-1281.
[10]罗毅,邓琼飞,杨昆,等.近20年来中国典型区域PM2.5时空演变过程[J].环境科学,2018,39(7):3003-3013.
[11]VESNA M,MARLEEN T-C,MICHEL V P.O20 Deep learning with convolutional neural networks for detection of interictal epileptiform discharges[J].Clinical Neurophysiology,2017,128(9):186.
[12]MARLEEN T-C,JEANNATTA H,BARRY R,et al.O77 Deep learning for EEG classification for outcome prediction of postanoxic coma[J].Clinical Neurophysiology,2017,128(9):203.
[13]SATO M,HORIE K,HARA A,et al.Application of deep learning to the classification of images from colposcopy[J].Oncology letters,2018,15(3):3518-3523.
[2]王晓彦,刘冰,李健军,等.区域环境空气质量预报的一般方法和基本原则[J].中国环境监测,2015,31(1):134-138.
[3]吴兑,邓雪娇,林爱兰,等.广东省空气质量预报系统[J].气象科技,2003(6):351-355.
[4]李小飞,张明军,王圣杰,等.中国空气污染指数变化特征及影响因素分析[J].环境科学,2012(6):1936-1943.
[5]朱玉强.几种空气质量预报方法的预报效果对比分析[J].气象,2004(10):30-33.
[6]陈彬彬,林长城,杨凯,等.基于CMAQ模式产品的福州市空气质量预报系统[J].中国环境科学,2012,32(10):1744-1752.
[7]任万辉,苏枞枞,赵宏德.城市环境空气污染预报研究进展[J].环境保护科学,2010(3):9-11.
[8]沈路路,王聿绚,段雷.神经网络模型在O3浓度预测中的应用[J].环境科学,2011,32(8):2231-2235.
[9]赵宏,刘爱霞,王恺,等基于GA-ANN改进的空气质量预测模型[J].环境科学研究,2009(11):1276-1281.
[10]罗毅,邓琼飞,杨昆,等.近20年来中国典型区域PM2.5时空演变过程[J].环境科学,2018,39(7):3003-3013.
[11]VESNA M,MARLEEN T-C,MICHEL V P.O20 Deep learning with convolutional neural networks for detection of interictal epileptiform discharges[J].Clinical Neurophysiology,2017,128(9):186.
[12]MARLEEN T-C,JEANNATTA H,BARRY R,et al.O77 Deep learning for EEG classification for outcome prediction of postanoxic coma[J].Clinical Neurophysiology,2017,128(9):203.
[13]SATO M,HORIE K,HARA A,et al.Application of deep learning to the classification of images from colposcopy[J].Oncology letters,2018,15(3):3518-3523.