本文提出一种基于多尺度时空优化的空气质量预测方法(multi-scale spatial-temporal network for air quality prediction,MSSTN-AQP),结合空气质量系统中存在的长短期时间依赖关系和动态空间依赖性,提高长期空气质量预测的准确性。首先...本文提出一种基于多尺度时空优化的空气质量预测方法(multi-scale spatial-temporal network for air quality prediction,MSSTN-AQP),结合空气质量系统中存在的长短期时间依赖关系和动态空间依赖性,提高长期空气质量预测的准确性。首先,通过构建多尺度时空特征提取模块,从多源异构数据中提取时空特征。其次,构建动态空间特征提取模块。通过将图卷积网络与注意力机制进行有效结合,捕捉空气质量网络中的全局空间特征,用于对多种空间依赖关系的联合建模。最后,构建时间特征提取模块,对Transformer模型进行改进与优化。自适应时间Transformer模块主要用于模拟跨多个时间步长的双向时间依赖关系。此外,将上述时空特征提取模块进行有效集成化,构建端到端的空气质量预测模型。为了验证模型的有效性,在两个真实数据集中进行实验验证。实验结果表明,MSSTN-AQP在预测精度上更具优势,尤其是在长期的空气质量预测任务中优势更加明显。展开更多
针对传统空气质量预测模型收敛速度慢,精度低的问题,提出一种基于变分模态分解(variational mode decomposition,VMD)和蜣螂优化算法(dung beetle optimizer,DBO)优化长短期记忆网络(long short term memory,LSTM)的预测模型。首先,针对...针对传统空气质量预测模型收敛速度慢,精度低的问题,提出一种基于变分模态分解(variational mode decomposition,VMD)和蜣螂优化算法(dung beetle optimizer,DBO)优化长短期记忆网络(long short term memory,LSTM)的预测模型。首先,针对AQI原始数据具有大量噪声的问题,使用VMD方法对非平稳信号进行模态分解以降低噪声对预测结果的影响从而获得多个不同特征的模态分量;其次,针对LSTM靠人工经验调参存在一定局限性,利用DBO算法对LSTM模型参数进行优化;最后,对分解后的各个子序列使用LSTM模型预测,将各个子序列进行叠加得到最后的预测结果。实验结果表明,VMD对非平稳数据的分解有助于提高预测精度,VMD-DBO-LSTM模型的性能较其他模型均有不同程度的提高,该模型预测的均方根误差为4.73μg/m^(3),平均绝对误差为3.61μg/m^(3),拟合度达到了97.8%。展开更多
空气质量预测是治理和减轻空气污染的有效手段。为了提高预测精度,构建了一个新的空气质量预测模型,即样本卷积和交互网络(sample convolutional and interaction network,SCINet)模型。该模型由多个SCIBlock按照完全二叉树结构排列而成...空气质量预测是治理和减轻空气污染的有效手段。为了提高预测精度,构建了一个新的空气质量预测模型,即样本卷积和交互网络(sample convolutional and interaction network,SCINet)模型。该模型由多个SCIBlock按照完全二叉树结构排列而成,通过翻转奇偶分裂重新排列生成一个新的序列,该结构能够更好地捕捉多变量大气污染物彼此间复杂的依赖关系和局部趋势。因为大气污染物监测数据具有季节性和随机性,所以使用两个SCINet进行叠加,既能扩大卷积运算的接受域,又能实现多分辨率分析。此外,通过模型深度及超参数调优,使其更加拟合空气质量时序数据特征,能够有助于提取目标变量的时间关系特征。最后,通过北京PM_(2.5)数据集和北京多站点空气质量数据集进行实证研究,结果表明,SCINet模型具有更高的预测精度,在短期预测中其均方根误差比对比模型中表现最佳的DAQFF模型减少了31.59%,在长期预测中减少了24.36%。展开更多
本文旨在通过构建EEMD-GWO-LSTM混合模型,对兰州市空气质量指数(AQI)进行准确预测。兰州市作为中国西北地区重要的工业基地和交通枢纽,其空气质量受工业排放、交通污染及地理环境等因素影响,常年处于高污染等级。针对兰州市AQI监测数据...本文旨在通过构建EEMD-GWO-LSTM混合模型,对兰州市空气质量指数(AQI)进行准确预测。兰州市作为中国西北地区重要的工业基地和交通枢纽,其空气质量受工业排放、交通污染及地理环境等因素影响,常年处于高污染等级。针对兰州市AQI监测数据突变性强的特点,文章首先对数据进行预处理,包括填补缺失值、归一化处理等,以提高数据质量。随后,采用集合经验模态分解(EEMD)对数据进行分解,提取出本征模态函数(IMF),并利用灰狼优化算法(GWO)对长短期记忆网络(LSTM)模型的超参数进行优化,以提高预测精度。实验结果表明,EEMD-GWO-LSTM混合模型在预测兰州市AQI时,相较于单一模型和其他混合模型,具有更低的均方根误差(RMSE)和更高的决定系数(R2),显示出更好的预测性能。最后,文章提出了增加监测站点、采用先进技术提高监测频率、跨区域合作及数据公开共享等建议,以促进兰州市空气质量的持续改善和预测模型的进一步优化。This paper aims to accurately predict the Air Quality Index (AQI) of Lanzhou City by constructing a mixed model of EEMD-GGO-LSTM. Lanzhou City is an important industrial base and transportation hub in northwest China. Its air quality is affected by industrial emissions, traffic pollution, and the geographical environment, and it always has a high pollution level. In view of the strong mutability of AQI monitoring data in Lanzhou City, the paper first preprocessed the data, including filling in missing values and normalization processing, so as to improve the data quality. Then, the data is decomposed by ensemble empirical Mode decomposition (EEMD), extracting the intrinsic mode function (IMF), and the hyperparameters of the long short-term memory network (LSTM) model are optimized by Grey Wolf optimization algorithm (GWO) to improve the prediction accuracy. The experimental results show that the EEMD-GGO-LSTM mixed model has a lower root-mean-square error (RMSE) and higher determination coefficient (R2) when predicting AQI in Lanzhou compared with the single model and other mixed models, showing better prediction performance. Finally, the paper puts forward some suggestions, such as increasing monitoring stations, using advanced technology to improve monitoring frequency, cross-regional cooperation and open data sharing, so as to promote the continuous improvement of Lanzhou air quality and further optimization of a prediction model.展开更多
本文依据往年河南省空气质量数据建立ARIMA模型以及BP神经网络模型,对河南省未来的空气质量进行预测研究,并比较其预测效果。研究构建ARIMA模型以及BP神经网络空气质量预测模型,经相关检验修正符合要求后,建模结果显示所建模型精度高,...本文依据往年河南省空气质量数据建立ARIMA模型以及BP神经网络模型,对河南省未来的空气质量进行预测研究,并比较其预测效果。研究构建ARIMA模型以及BP神经网络空气质量预测模型,经相关检验修正符合要求后,建模结果显示所建模型精度高,可以满足对河南省空气质量预测要求,具有实践意义。研究结果表明,ARIMA模型预测结果显示2023年河南省AQI仍呈现季节性变化但相较于2018~2022年有所下降。BP模型预测结果显示,2023年河南省AQI逐月变化呈“V”形,据年变化而言,河南省空气质量在向好的趋势发展。ARIMA模型和BP神经网络模型均能有效预测河南省的空气质量,但BP模型的拟合效果以及均方误差均优于ARIMA模型,预测曲线与真实曲线更接近,均方误差更小,因此可为河南省空气质量预测提供依据。Based on the air quality data of Henan Province in previous years, ARIMA model and BP neural network model are established in this paper to predict the future air quality of Henan Province, and compares their prediction effects. Research and construction of ARIMA model and BP neural network air quality prediction model. After relevant testing and correction, the modeling results show that the established model has high accuracy and can meet the requirements of air quality prediction in Henan Province, which has practical significance. The research results indicate that the ARIMA model predicts that the AQI in Henan Province will still show seasonal changes in 2023, but will decrease compared to 2018~2022. The BP model prediction results show that the monthly AQI in Henan Province will change in a “V” shape in 2023. According to the annual changes, the air quality in Henan Province is developing in a positive trend. Both ARIMA model and BP neural network model can effectively predict the air quality in Henan Province, but the fitting effect and mean square error of BP model are better than ARIMA model. The predicted curve is closer to the real curve, and the mean square error is smaller. Therefore, it can provide a basis for predicting the air quality in Henan Province.展开更多
文摘本文提出一种基于多尺度时空优化的空气质量预测方法(multi-scale spatial-temporal network for air quality prediction,MSSTN-AQP),结合空气质量系统中存在的长短期时间依赖关系和动态空间依赖性,提高长期空气质量预测的准确性。首先,通过构建多尺度时空特征提取模块,从多源异构数据中提取时空特征。其次,构建动态空间特征提取模块。通过将图卷积网络与注意力机制进行有效结合,捕捉空气质量网络中的全局空间特征,用于对多种空间依赖关系的联合建模。最后,构建时间特征提取模块,对Transformer模型进行改进与优化。自适应时间Transformer模块主要用于模拟跨多个时间步长的双向时间依赖关系。此外,将上述时空特征提取模块进行有效集成化,构建端到端的空气质量预测模型。为了验证模型的有效性,在两个真实数据集中进行实验验证。实验结果表明,MSSTN-AQP在预测精度上更具优势,尤其是在长期的空气质量预测任务中优势更加明显。
文摘针对传统空气质量预测模型收敛速度慢,精度低的问题,提出一种基于变分模态分解(variational mode decomposition,VMD)和蜣螂优化算法(dung beetle optimizer,DBO)优化长短期记忆网络(long short term memory,LSTM)的预测模型。首先,针对AQI原始数据具有大量噪声的问题,使用VMD方法对非平稳信号进行模态分解以降低噪声对预测结果的影响从而获得多个不同特征的模态分量;其次,针对LSTM靠人工经验调参存在一定局限性,利用DBO算法对LSTM模型参数进行优化;最后,对分解后的各个子序列使用LSTM模型预测,将各个子序列进行叠加得到最后的预测结果。实验结果表明,VMD对非平稳数据的分解有助于提高预测精度,VMD-DBO-LSTM模型的性能较其他模型均有不同程度的提高,该模型预测的均方根误差为4.73μg/m^(3),平均绝对误差为3.61μg/m^(3),拟合度达到了97.8%。
文摘空气质量预测是治理和减轻空气污染的有效手段。为了提高预测精度,构建了一个新的空气质量预测模型,即样本卷积和交互网络(sample convolutional and interaction network,SCINet)模型。该模型由多个SCIBlock按照完全二叉树结构排列而成,通过翻转奇偶分裂重新排列生成一个新的序列,该结构能够更好地捕捉多变量大气污染物彼此间复杂的依赖关系和局部趋势。因为大气污染物监测数据具有季节性和随机性,所以使用两个SCINet进行叠加,既能扩大卷积运算的接受域,又能实现多分辨率分析。此外,通过模型深度及超参数调优,使其更加拟合空气质量时序数据特征,能够有助于提取目标变量的时间关系特征。最后,通过北京PM_(2.5)数据集和北京多站点空气质量数据集进行实证研究,结果表明,SCINet模型具有更高的预测精度,在短期预测中其均方根误差比对比模型中表现最佳的DAQFF模型减少了31.59%,在长期预测中减少了24.36%。
文摘本文旨在通过构建EEMD-GWO-LSTM混合模型,对兰州市空气质量指数(AQI)进行准确预测。兰州市作为中国西北地区重要的工业基地和交通枢纽,其空气质量受工业排放、交通污染及地理环境等因素影响,常年处于高污染等级。针对兰州市AQI监测数据突变性强的特点,文章首先对数据进行预处理,包括填补缺失值、归一化处理等,以提高数据质量。随后,采用集合经验模态分解(EEMD)对数据进行分解,提取出本征模态函数(IMF),并利用灰狼优化算法(GWO)对长短期记忆网络(LSTM)模型的超参数进行优化,以提高预测精度。实验结果表明,EEMD-GWO-LSTM混合模型在预测兰州市AQI时,相较于单一模型和其他混合模型,具有更低的均方根误差(RMSE)和更高的决定系数(R2),显示出更好的预测性能。最后,文章提出了增加监测站点、采用先进技术提高监测频率、跨区域合作及数据公开共享等建议,以促进兰州市空气质量的持续改善和预测模型的进一步优化。This paper aims to accurately predict the Air Quality Index (AQI) of Lanzhou City by constructing a mixed model of EEMD-GGO-LSTM. Lanzhou City is an important industrial base and transportation hub in northwest China. Its air quality is affected by industrial emissions, traffic pollution, and the geographical environment, and it always has a high pollution level. In view of the strong mutability of AQI monitoring data in Lanzhou City, the paper first preprocessed the data, including filling in missing values and normalization processing, so as to improve the data quality. Then, the data is decomposed by ensemble empirical Mode decomposition (EEMD), extracting the intrinsic mode function (IMF), and the hyperparameters of the long short-term memory network (LSTM) model are optimized by Grey Wolf optimization algorithm (GWO) to improve the prediction accuracy. The experimental results show that the EEMD-GGO-LSTM mixed model has a lower root-mean-square error (RMSE) and higher determination coefficient (R2) when predicting AQI in Lanzhou compared with the single model and other mixed models, showing better prediction performance. Finally, the paper puts forward some suggestions, such as increasing monitoring stations, using advanced technology to improve monitoring frequency, cross-regional cooperation and open data sharing, so as to promote the continuous improvement of Lanzhou air quality and further optimization of a prediction model.
文摘本文依据往年河南省空气质量数据建立ARIMA模型以及BP神经网络模型,对河南省未来的空气质量进行预测研究,并比较其预测效果。研究构建ARIMA模型以及BP神经网络空气质量预测模型,经相关检验修正符合要求后,建模结果显示所建模型精度高,可以满足对河南省空气质量预测要求,具有实践意义。研究结果表明,ARIMA模型预测结果显示2023年河南省AQI仍呈现季节性变化但相较于2018~2022年有所下降。BP模型预测结果显示,2023年河南省AQI逐月变化呈“V”形,据年变化而言,河南省空气质量在向好的趋势发展。ARIMA模型和BP神经网络模型均能有效预测河南省的空气质量,但BP模型的拟合效果以及均方误差均优于ARIMA模型,预测曲线与真实曲线更接近,均方误差更小,因此可为河南省空气质量预测提供依据。Based on the air quality data of Henan Province in previous years, ARIMA model and BP neural network model are established in this paper to predict the future air quality of Henan Province, and compares their prediction effects. Research and construction of ARIMA model and BP neural network air quality prediction model. After relevant testing and correction, the modeling results show that the established model has high accuracy and can meet the requirements of air quality prediction in Henan Province, which has practical significance. The research results indicate that the ARIMA model predicts that the AQI in Henan Province will still show seasonal changes in 2023, but will decrease compared to 2018~2022. The BP model prediction results show that the monthly AQI in Henan Province will change in a “V” shape in 2023. According to the annual changes, the air quality in Henan Province is developing in a positive trend. Both ARIMA model and BP neural network model can effectively predict the air quality in Henan Province, but the fitting effect and mean square error of BP model are better than ARIMA model. The predicted curve is closer to the real curve, and the mean square error is smaller. Therefore, it can provide a basis for predicting the air quality in Henan Province.
