基于BP神经网络的大气污染物浓度预测

利用BP神经网络结合变量筛选的方法建立了SO_2、NO_2、O3、CO、PM_(10)、PM_(2.5)等6种污染物的浓度预测模型,并选取2014-01-01至2015-11-28时段,昆明市区5个环境监测点以上6种污染物浓度的监测数据建立了昆明市污染物日均浓度预测模型.采用平均影响值(Mean Impact Value,MIV)的方法筛选出分别对6种污染物日均浓度值有主要影响的变量,作为BP神经网络的输入变量,利用建立的预测模型分别对6种污染物的日均浓度进行预测.结果表明,在关上监测点利用浓度预测模型对SO_2、NO_2、O3、CO、PM_(10)、PM_(2.5)等6种污染物浓度进行预测,污染物浓度预测值和实测值趋势吻合度较高.变量筛选后SO_2、PM_(2.5)预测效果比变量筛选前的预测效果好.O3的均方根误差和PM_(10)的标准化平均偏差,变量筛选前的预测效果比变量筛选后的预测效果好.变量筛选前的NO_2和CO的预测结果比变量筛选后的预测效果好.其他4个环境监测点的污染物浓度预测结果与关上监测点的结果相似.     

Polychlorinated biphenyls in the atmosphere of Taizhou, a major e-waste dismantling area in China

PM2.5,total suspended particles(TSP) and gas phase samples were collected at two sites of Taizhou,a major e-waste dismantling area in China.Concentrations,seasonal variations,congener profiles,gas-particle partitioning and size distribution of the atmospheric polychlorinated biphenyls(PCBs) were studied to assess the current state of atmospheric PCBs after the phase out of massive historical dismantling of PCBs containing e-wastes.The average 38PCBs concentration in the ambient air(TSP plus gas phase) near the e-waste dismantling area was(12,407 ± 9592) pg/m3 in winter,which was substantially lower than that found one decade ago.However,the atmospheric PCBs level near the e-waste dismantling area was 54 times of the reference urban site,indicating that the impact of the historical dismantling of PCBs containing e-wastes was still significant.Tri-Penta-CBs were dominant homologues,consisting with their dominant global production.Size distribution of particle-bound PCBs showed that higher chlorinated CBs tended to partition more to the fine particles,facilitating its long range air transportation.     

新兴污染物BP-3和BP-4的好氧生物降解性能

采用欧洲经济合作与发展组织（OECD）的生物降解测试标准方法——301F测压呼吸计量法,考察了2-羟基-4-甲氧基二苯甲酮（BP-3）和2-羟基-4-甲氧基二苯甲酮-5-磺酸（BP-4）的好氧生物降解性能,并研究了降解动力学及共代谢现象。实验结果表明：BP-3和BP-4的可生物降解率分别为68.36%和41.34%;根据OECD快速降解性判定标准,BP-3划归为易快速降解物质,而BP-4为不易快速降解物质;两种物质的生物降解可用一级动力学描述,半衰期分别为1.986 d和2.806 d;根据欧盟法规《化学品的注册、评估、授权和限制》（REACH法规）,BP-3和BP-4均非持久性物质;与苯甲酸钠共存时,BP-3和BP-4的降解过程均表现出共代谢现象。     

水质的生化需氧量(BOD5)稀释与接种法测定结果不确定度评定

在实际工作中，运用测量过程的合并样本标准差来评定A类不确定度比较客观。本文根据JJF1059—1999《测量不确定度评定与表示》，通过实例，阐述了用稀释接种法测定水中生化需氧量不确定度的评定方法。     

甲氰菊酯降解菌JQL4-5的分离鉴定及降解特性研究

从农药厂废水处理池的活性污泥中分离到1株能以甲氰菊酯为唯一碳源生长的细菌,命名为JQL4-5.根据其生理生化特征和16S rDNA(GenBank Accession No.DQ177525)序列相似性分析,将该菌株鉴定为鞘氨醇单胞菌属(Sphingomonas sp.).该菌株在24h内对20 mg/L的甲氰菊酯的降解率达到99.8%.降解甲氰菊酯的最适温度为30℃,pH为7.0,降解速率与初始接种量呈正相关.酶的定域试验表明,降解甲氰菊酯的酶为胞内酶.     

基于远程在线监控车载终端的柴油车NOx排放等级诊断研究

远程在线监控车载终端集成了远程通讯模块、卫星定位模块、发动机OBD信息解析模块,能够实时读取车辆排放相关运行信息,但无法直接判断车辆NO_x排放情况.为了快速、准确地评估车辆排放情况,诊断和监测NO_x高排放车,同时为了克服有些重型柴油车监测数据中缺失进气流量、燃油流量、车速等重要的实时信息,无法计算出车辆NO_x排放因子的问题.本文提出了由NO_x浓度分布特征驱动的高排放重型柴油车识别算法,通过远程在线监控车载终端设备获取车辆的发动机信息和SCR系统运行信息,运用NO_x浓度分布计算车辆每天NO_x排放浓度占比,通过系统聚类法对车辆NO_x排放浓度占比进行聚类,结果聚为优、良、中、差4类.利用车辆NO_x排放浓度区间分布及其聚类结果分别作为训练集的输入和输出,选择BP神经网络作为训练算法,训练获得的模型分类准确率为90%,利用训练好的模型判断在用柴油车NO_x排放等级,从而识别及监测NO_x高排放车辆.研究结果可为柴油车NO_x高排放诊断及监测提供依据,有助于监管部门能够快速识别NO_x高排放车辆.     

Characteristics and source identification of fine particulate n-alkanes in Beijing, China

Ambient particulate n-alkanes were determined for fine particle (PM2:5) samples collected from Sep 2003 to July 2004 in Beijing, China. The average concentration of total n-alkanes (Pn-alkanes) from C11 to C34 was 425.72 ng/m3, ranged from 7.02 to 2893.28 ng/m3. The concentration distributions of n-alkanes homologues in this study exhibited peaks at C21 and C29 in heating season, and C29 in non-heating season. The average carbon preference index (CPI) value was 1.88 in the range of 1.18–3.88. The maximum CPI in summer indicated the contribution of biogenic origins such as plant wax; while the minimum CPI value in winter was probably a result of fossil fuel combustion. Preliminary estimation from these results showed that 59% of the n-alkanes in PM2:5 in Beijing summer originated from plant wax, while 74%–88% was from fossil fuel combustion in other three seasons. Source estimation was further performed using principal component analysis method. Two major components were yielded accounting for 57.3% and 30.9% of the total variance, which presented the fossil fuel and biogenic contribution, respectively.     

基于BP神经网络的城市总生态足迹预测研究——以武汉市为例

鉴于BP网络在处理非线性复杂系统的优势,以武汉市为研究对象,构建一个10-4-1结构的BP神经网络预测模型,将1978-2002年和2003年的相关数据作为模型的训练和测试样本,以2004年的社会、经济、环境数据作为网络的预测输入,对该年的总生态足迹进行预测。结果表明,BP神经网络预测结果与实际足迹值的相对误差为0.69%,预测精度优于传统的多元回归统计模型。     

Developing chemical signatures of particulate air pollution in the Pearl River Delta region, China

PM2:5 samples were collected in a regional sampling network with three sites in Hong Kong and four sites in the adjacent inland Pearl River Delta (PRD) or Guangdong Province during four months/seasons from 2002–2003. Trans-boundary transport between Hong Kong and the inland PRD is inevitable under the influence of Asian monsoon. In summer, Hong Kong serves as the upwind site of the inland PRD while during other seasons it is under the influence of continental emissions. Previous studies have recognized the importance of using chemical signatures to differentiate local vs. regional contributions to air pollutants in Hong Kong such as the CO/NOx ratio, ratios of different VOC species. In this study, detailed chemical speciation by gas chromatography-mass spectrometry was performed with PM2:5 samples to identify new chemical signatures to distinguish aerosols in Hong Kong from those from the inland PRD. Since Hong Kong is not influenced by the continental emissions from the inland PRD during summer, comparison focused on chemical data obtained from this season for chemical signatures. The new ratios developed from the current study include LCPI/HCPI ratio of alkanes (0.39 0.02 in Hong Kong vs. 0.78 0.08 in the inland PRD), pyrene to benzo[ghi]perylene ratio (0.97 0.21 in Hong Kong compared to 0.20 0.06 in the inland PRD), and the ratio of 1,2-benzenedioic acid to 1,4-benzenedioic acid (1.8 0.1 in Hong Kong vs. 0.6 0.05 in the inland PRD). Results from this study also revealed that Hong Kong was impacted by ship emissions as reflected by substantially high V/Ni ratio (9 2) while this ratio was about 1–2 at all sites in the inland PRD, which is very close to typical ratios from residual oil combustion.     

基于BP神经网络的污泥水解液合成PHA的多参数敏感性分析

本文采用多参数敏感性分析方法对影响污泥水解液合成PHA产量的参数进行分析.在实验数据的基础上,利用BP神经网络建立了PHA的产量预测模型.通过与真实试验结果的对比,验证了预测模型的精确度.根据训练完成的神经网络模型中的各参数变量到目标的权值和阈值,利用Garson算法定量得到各参数变量对于目标的参数敏感性系数数值.结论表明：基于BP神经网络技术建立的预测模型具有较高的可信度,多参数敏感性分析方法可评估多因素同时变化对PHA产量的影响,具有较高的实用价值.