基于KNN-LSTM的PM2.5浓度预测模型

目前多数PM_2.5浓度预测模型仅利用单个站点的时间序列数据进行浓度预测, 并没有考虑到空气质量监测站之间的区域关联性, 这会导致预测存在一定的片面性. 本文利用KNN算法选择目标站点所在区域中与其相关的空间因素, 并结合LSTM模型, 提出基于时空特征的KNN-LSTM的PM_2.5浓度预测模型. 以哈尔滨市10个空气质量监测站的污染物数据进行仿真实验, 并将KNN-LSTM模型与其他预测模型进行对比, 结果显示: 模型相较于BP神经网络模型平均绝对误差(MAE)、均方根误差(RMSE)分别降低了19.25%、13.23%; 相较于LSTM模型MAE、RMSE分别降低了4.29%、6.99%. 表明本文所提KNN-LSTM模型能有效提高LSTM模型的预测精度.     

基于自适应层级图卷积的多站点空气质量预测模型

时空预测任务在污染治理、交通、能源、气象等领域应用广泛. PM_2.5浓度预测作为典型的时空预测任务, 需要对空气质量数据中的时空依赖关系进行分析和利用. 现有时空图神经网络(ST-GNNs)研究所使用的邻接矩阵使用启发式规则预定义, 无法准确表示站点之间的真实关系. 本文提出了一种自适应分层图卷积神经网络(AHGCNN)用于PM_2.5预测. 首先, 引入了一种分层映射图卷积架构, 在不同层级上使用不同的自学习邻接矩阵, 以有效挖掘不同站点之间独特的时空依赖. 其次, 以基于注意力的聚合机制连接上下层邻接矩阵, 加速收敛过程. 最后, 将隐藏的空间状态与门控循环单元相结合, 形成一个统一的预测架构, 同时捕捉多层次的空间依赖关系和时间依赖关系, 提供最终的预测结果. 实验中, 我们与7种主流预测模型进行对比, 结果表明该模型可以有效获取空气监测站点之间的时空依赖, 提高预测精确度.     

基于Spring Boot的云端数据监控管理与可视化应用系统

为了实现对存储在云端空气质量数据的管理与实时可视化展示,构建了一个基于Spring Boot的云端数据监控与可视化系统.系统为B/S架构,采用Spring Boot框架搭建后端微服务实例,使系统的配置与监控变得简单.Vue.js框架实现前端页面开发.使用Axios插件封装的Ajax技术来实现数据交互,实现前后端连接与逻辑交互同时,减少服务器开销与响应.本系统通过云端数据库可以查询到11种空气成分信息,包括：PM_1.0、PM_2.5、PM₁₀、CO、CO₂、NO、NO₂、O₃、SO₂、甲醛、TVOC,除此之外还监测包括温度、湿度、风速、坐标,时间等其他相关属性.实现了数据监控下载、报警管理、百度地图可视化多个功能.程序将部署在阿里云端,方便用户远程访问web项目.其中Spring Boot框架与Vue的结合实现了系统的前后端分离,使系统具有良好的稳定性、实时性与高效性.     

基于最优定权组合法的大气污染物SO2预测

为了提高对大气污染物SO₂的预测准确率,基于多个空气质量预测模式（WRF-CHEM、CMAQ、CAMx）,以过去一段时间内各单项空气质量预测模式的组合预测误差平方和最小为原则,构建出针对大气污染物SO₂的最优定权组合预测模型.选取2018年云南省楚雄、昭通、蒙自三个站点1至5月份的实际观测数据和前述三个空气质量模式的预测数据作为实验样本,然后分别采用多元线性回归法和动态权重更新法在相同的实验条件下与所提的最优定权组合预测法进行预测对比实验.实验结果表明,所提方法的预测值相较其他两种方法更加贴近实际观测值,其两项误差评估指标值均最小.总体而言,最优定权组合预测模型很好地综合了各单项空气质量预测模式的优势,提高了SO₂的预测精度.     

GA-ELM在硫铁矿制酸尾气SO2浓度预测的应用

基于对整个生产流程的管控,使硫铁矿生产硫酸尾气的SO₂浓度达标排放,提出运用GA-ELM对制酸尾气SO₂浓度进行建模预测.在硫铁矿制酸的生产过程中采集对尾气SO₂浓度影响较大的关键点参数,运用GA-ELM神经网络对烟气制酸尾气SO₂浓度进行预测.该方法在某厂实际检验,其预测结果与实际数据吻合度较高,对于调整和优化工艺指标和尾气达标排放起到很好的指导作用.     

基于多元线性回归的雾霾预测方法研究

提出了一种在线样本更新的多元线性回归分析的雾霾预测方法。首先搜集了北京市天气状况,包括平均气温、湿度、风级等气象数据以及PM2.5、CO、NO₂、SO₂等大气成分浓度数据,然后通过散点图对这些因素进行主要影响因素分析,筛选出对雾霾影响比较明显的因素作为雾霾预测的依据。通过在线样本更新的多元线性回归建立了PM2.5含量预测模型,并将气象要素作为雾霾的判断标准。最后给出实际例子,利用多元线性回归对北京未来一天、三天及一周的PM2.5含量进行较为精确的预测。     

混合H2/H∞鲁棒控制器设计

在状态空间描述下,定义了混合H₂/H_∞控制的完整信息、完整控制、干扰顺馈、输出估计这4种典型情况.在二次稳定意义上,讨论了混合H₂/H_∞的性能指标,及这4种典型情况的混合H₂/H_∞线性反馈控制器设计,给出了充分必要条件.在典型情况分析的基础上,研究一般意义上的混合H₂/H_∞反馈控制器设计.H₂和H_∞的干扰输入阵及性能评价函数各不相同时的混合H₂/H_∞反馈控制器,与H₂和H_∞控制器设计相似,归结为解两个Riccati方程.但这两个Riccati方程含有参数,最优解要通过搜索这两个参数得到.结果包含了单纯的H₂和H_∞设计,可看作是H₂,H_∞和混合H₂/H_∞的统一设计方法.最后通过一个简单的例子,说明了控制器设计方法的可行性.     

基于多特征融合的航空发动机剩余寿命预测

针对航空发动机剩余可用寿命(RUL)预测任务中代表性特征提取不充分导致RUL预测精度较低等问题, 提出了一种基于多特征融合的航空发动机RUL预测方法. 利用指数平滑法(ES)降低原始数据中的噪声干扰, 得到相对平稳的特征数据. 使用双向长短期记忆网络(Bi-LSTM)提取特征数据的时序特征, 利用多头注意力机制(Multi-attention)为时序特征赋予权重; 设计卷积长短期记忆网络(Conv-LSTM)提取特征数据的时空特征; 提取特征数据的手工特征并使用Softmax函数计算权重. 设计一个特征融合框架将上述特征进行融合, 然后通过全连接网络回归实现最终RUL预测. 使用C-MAPSS数据集对模型进行仿真验证, 与Bi-LSTM等模型进行对比, 模型RUL预测精度更高, 适应性更好.     

基于模糊时序和支持向量机的高速公路SO2浓度预测算法

针对现有SO₂浓度预测方法中存在的污染物来源和影响因素认识不统一、小样本数据敏感、易于陷入局部最优等问题,文中提出了基于模糊时序和支持向量机的高速公路SO₂浓度预测算法,为搭建高速公路环境健康监测系统提供了可靠的理论支持.该方法依据SO₂浓度的季节变动规律,以季节作为时间序列,以24h为粒化窗宽,通过高斯核函数提取原始样本数据的特征值,输入支持向量机训练模型,并利用k重交叉验证法结合网格划分优化模型参数.文中应用该方法建立了SO₂浓度预测模型,并以2014年4月至2015年3月山西省太旧高速公路某监测点SO₂小时浓度监测值为样本数据,在MATLAB平台下应用LIBSVM工具实现了计算过程.结果表明,基于模糊时序和支持向量机的高速公路SO₂浓度预测算法不受机理性理论研究的限制,支持小样本学习,非线性拟合效果好,泛化能力强.     

混合H2/H∞参数辨识

研究了混合H₂/H_∞参数辨识问题.将混合H₂/H_∞估计方法应用到系统参数辨识中,给出了混合H₂/H_∞参数辨识算法.所得的算法不仅能满足规定的鲁棒性能,且为最小二乘(LS)参数估计误差判据提供了一个最优上界.结果表明:提高辨识的鲁棒性,需要牺牲辨识的精度作为代价.最后,仿真结果也验证了该方法的有效性.     

基于GA-GRU环境空气污染物预测研究

针对现有的环境空气污染物预测方法存在缺少输入特征相关性分析和时序信息丢失问题,提出一种将遗传算法和门控循环单元神经网络相结合的环境空气污染物PM2.5小时浓度预测模型,充分挖掘了污染物时间序列内在依赖关系,并解决了不相关因子的干扰和输入特征维度灾难的问题。最后基于绵阳市4个空气污染物监测站点的数据集进行仿真实验,与门控循环单元神经网络、深度信念网络预测比较。结果表明,基于GA-GRU的PM2.5小时浓度预测模型在训练时间、预测精度和鲁棒性上优势显著,是一种可行且有效的预测方法。     

A semi-empirical model for predicting hourly ground-level fine particulate matter (PM2.5) concentration in southern Ontario from satellite remote sensing and ground-based meteorological measurements

A semi-empirical model is developed to predict the hourly concentration of ground-level fine particulate matter (PM_2.5) coincident to satellite overpass, at a regional scale. The model corrects the aerosol optical depth (AOD) data from the Moderate Resolution Imaging Spectroradiometer (MODIS) by the assimilated parameters characterizing the boundary layer and further adjusts the corrected value according to meteorological conditions near the ground. The model was built and validated using the data collected for southern Ontario, Canada for 2004. Overall, the model is able to explain 65% of the variability in ground-level PM_2.5 concentration. The model-predicted values of PM_2.5 mass concentration are highly correlated with the actual observations. The root-mean-square error of the model is 6.1 µg/m³. The incorporation of ground-level temperature and relative humidity is found to be significant in improving the model predictability. The coarse resolution of the assimilated meteorological fields limits their value in the AOD correction. Although MODIS AOD data is acquired on a daily basis and the valid data coverage can sometimes be very limited due to unfavourable weather conditions, the model provides a cost-effective approach for obtaining supplemental PM_2.5 concentration information in addition to the ground-based monitoring station measurement.     

An assessment of air-quality monitoring station locations based on satellite observations

ABSTRACT

Optimization of the locations of air quality monitoring stations has great importance in providing high-quality data for regional air pollution monitoring. To assess the representativeness of the locations of the current air quality monitoring stations, we propose a new method based on satellite observations by applying the stratified sampling approach. Unlike the traditional method, which relies on the simulated spatial distribution of air pollutants from dispersion models, we obtained the sampling population through observations from remote sensing. As a first step, the spatial distribution of aggregated air quality was obtained based on ground concentrations of particulate matter (aerodynamic diameters of less than 10 μm, PM₁₀), fine particulate matter (aerodynamic diameters of less than 2.5 μm, PM_2.5), nitrogen dioxide (NO₂), and sulphur dioxide (SO₂) derived from satellite observations. Second, the representativeness of locations of air quality monitoring stations was assessed using the stratified sampling method. The results demonstrated that air quality monitoring stations in Beijing-Tianjin-Hebei were clustered in areas with heavily polluted air, whereas the number of air quality monitoring stations was insufficient in areas with higher air quality. After optimization, the minimum relative error was only 6.77%. It is indicated that combing remote-sensing data with the stratified sampling approach has great potential in assessing the spatial representativeness of air quality monitoring stations.     

Unraveling environmental justice in ambient PM2.5 exposure in Beijing: A big data approach

Air pollution imposes significant environmental and health risks worldwide and is expected to deteriorate in the coming decade as cities expand. Measuring population exposure to air pollution is crucial to quantifying risks to public health. In this work, we introduce a big data analytics framework to model residents' stay and commuters' travel exposure to outdoor PM_2.5 and evaluate their environmental justice, with Beijing as an example. Using mobile phone and census data, we first infer travel demand of the population to derive residents' stay activities in each analysis zone, and then focus on commuters and estimate their travel routes with a traffic assignment model. Based on air quality observations from monitoring stations and a spatial interpolation model, we estimate the outdoor PM_2.5 concentrations at a 500-m grid level and map them to road networks. We then estimate the travel exposure for each road segment by multiplying the PM_2.5 concentration and travel time spent on the road. By combining the estimated PM_2.5 exposure and housing price harnessed from online housing transaction platforms, we discover that in the winter, Beijing commuters with low wealth level are exposed to 13% more PM_2.5 per hour than those with high wealth level when staying at home, but exposed to less PM_2.5 by 5% when commuting the same distance (due to lighter traffic congestion in suburban areas). We also find that the residents from the southern suburbs of Beijing have both lower level of wealth and higher stay- and travel- exposure to PM_2.5, especially in the winter. These findings inform more equitable environmental mitigation policies for future sustainable development in Beijing. Finally, or the first time in the literature, we compare the results of exposure estimated from passive data with subjective measures of perceived air quality (PAQ) from a survey. The PAQ data was collected via a mobile-app. The comparison confirms consistencies in results and the advantages of the big data for air pollution exposure assessments.     

Modelling PM2.5 impact indicators in Europe: Health effects and legal compliance

Long-term exposure to fine particulate matter (PM_2.5) has been shown to have significant negative impacts on human health. It is estimated that current levels of air pollution shorten the statistical life expectancy of European citizens by several months. The GAINS integrated assessment model calculates shortening of life expectancy from population exposure to PM_2.5 using epidemiologically-derived health impact functions. In addition, GAINS estimates PM_2.5 concentrations at 1875 air quality monitoring stations located in diverse environments ranging from remote background locations to busy street canyons. In this article, different approaches to dealing with the PM_2.5 pollution problem are compared. We assess for the present and future the attainment of EU and WHO air quality standards for PM_2.5 and estimate the loss of life expectancy under different policy scenarios developed for the ongoing revision of the EU Air Quality Legislation.     

Estimating and source analysis of surface PM2.5 concentration in the Beijing–Tianjin–Hebei region based on MODIS data and air trajectories

Particulate matter (PM) with an aerodynamic diameter of <2.5 μm (PM_2.5) has become the primary air pollutant in most major cities in China. Some studies have indicated that there is a positive correlation between the aerosol optical thickness (AOT) and surface-level PM_2.5 concentration. In order to estimate PM_2.5 concentration over large areas, a model relating the concentration of PM_2.5 and AOT has been established. The scale height of aerosol and relative humidity as well as the effect of surface temperature and wind velocity were introduced to enhance the model. 2013 full year Moderate Resolution Imaging Spectroradiometer (MODIS) AOT data and ground measurements of the PM_2.5 concentration in the Beijing–Tianjin–Hebei region were used to fit a seasonal multivariate linear equation relating PM_2.5 concentration and AOT, and the accuracy of the model has been determined. When comparing MODIS-estimated PM_2.5 with the measurements from ground monitoring stations during spring, summer, autumn and winter, we found the R² values were 0.45, 0.45, 0.37, and 0.31, respectively. Based on this model, the spatial distribution of PM_2.5 concentration during four typical haze events sampled by seasons was derived, and displayed with the backward air trajectories calculated using the Hybrid Single Particle Lagrangian Integrated Trajectory (HYSPLIT) model. We undertook a preliminary analysis about the source of surface-level PMs and the process of its accumulation and dispersion during the haze episodes by analysing the effect of terrain and topography in the specific location of the Beijing–Tianjin–Hebei region. The spatial distribution of the PM_2.5 concentration showed that the high value region was generally in the southeast of the study area, which approximately overlapped an area of lower vegetation coverage, and the temporal variation of PM_2.5 concentration indicated that the air pollution was more severe during winter and spring than summer and autumn. The results of the analysis of backward air trajectories suggested that the hazy weather in the Beijing–Tianjin–Hebei region was mainly caused by unfavourable terrain and weather conditions.     

PM2.5 concentration prediction using hidden semi-Markov model-based times series data mining

In this paper, a novel framework and methodology based on hidden semi-Markov models (HSMMs) for high PM_2.5 concentration value prediction is presented. Due to lack of explicit time structure and its short-term memory of past history, a standard hidden Markov model (HMM) has limited power in modeling the temporal structures of the prediction problems. To overcome the limitations of HMMs in prediction, we develop the HSMMs by adding the temporal structures into the HMMs and use them to predict the concentration levels of PM_2.5. As a model-driven statistical learning method, HSMM assumes that both data and a mathematical model are available. In contrast to other data-driven statistical prediction models such as neural networks, a mathematical functional mapping between the parameters and the selected input variables can be established in HSMMs. In the proposed framework, states of HSMMs are used to represent the PM_2.5 concentration levels. The model parameters are estimated through modified forward–backward training algorithm. The re-estimation formulae for model parameters are derived. The trained HSMMs can be used to predict high PM_2.5 concentration levels. The validation of the proposed framework and methodology is carried out in real world applications: prediction of high PM_2.5 concentrations at O’Hare airport in Chicago. The results show that the HSMMs provide accurate predictions of high PM_2.5 concentration levels for the next 24 h.     

Spatial-temporal association between fine particulate matter and daily mortality

Fine particulate matter (PM_2.5) is a mixture of pollutants that has been linked to serious health problems, including premature mortality. Since the chemical composition of PM_2.5 varies across space and time, the association between PM_2.5 and mortality could also change with space and season. A statistical multi-stage Bayesian framework is developed and implemented, which provides a very broad and flexible approach to studying the spatiotemporal associations between mortality and population exposure to daily PM_2.5 mass, while accounting for different sources of uncertainty. The first stage of the framework maps ambient PM_2.5 air concentrations using all available monitoring data (IMPROVE and FRM) and an air quality model (CMAQ) at different spatial and temporal scales. The second stage of the framework examines the spatial temporal relationships between the health end-points and the exposures to PM_2.5 by introducing a spatial-temporal generalized Poisson regression model. A method to adjust for time-varying confounders such as seasonal trends is proposed. A common seasonal trends model uses a fixed number of basis functions to account for these confounders, but the results can be sensitive to the number of basis functions. Thus, instead the number of the basis functions is treated as an unknown parameter in the Bayesian model, and a space-time stochastic search variable selection approach is used. The framework is illustrated using a data set in North Carolina for the year 2001.