Pipeline leakage detection and isolation: An integrated approach of statistical and wavelet feature extraction with multi-layer perceptron neural network (MLPNN)

Leakage diagnosis of hydrocarbon pipelines can prevent environmental and financial losses. This work proposes a novel method that not only detects the occurrence of a leakage fault, but also suggests its location and severity. The OLGA software is employed to provide the pipeline inlet pressure and outlet flow rates as the training data for the Fault Detection and Isolation (FDI) system. The FDI system is comprised of a Multi-Layer Perceptron Neural Network (MLPNN) classifier with various feature extraction methods including the statistical techniques, wavelet transform, and a fusion of both methods. Once different leakage scenarios are considered and the preprocessing methods are done, the proposed FDI system is applied to a 20-km pipeline in southern Iran (Goldkari-Binak pipeline) and a promising severity and location detectability (a correct classification rate of 92%) and a low False Alarm Rate (FAR) were achieved.     

高浓度氨氮废水的短程硝化研究

采用普通好氧活性污泥驯化培养启动亚硝化反应器，探索了在实验室条件下，亚硝化反应的最适宜条件。结果表明．在温度（T）为35℃，pH值为7．5左右，初始污泥浓度0．7g／L时，控制较高的初始进水氨氨浓度和较低的DO浓度，有利于亚硝化反应的启动：驯化后，反应器内氨氮处理效果良好，即使进水氨氮浓度高达2400mg／L时，氨氮去除率也能达到95％以上;在实验中，亚硝化的最适宜条件为，温度：29～35℃．pH值：7．0～8．0。同时．实验结果表明，在一定范围内，溶解氧浓度越高，亚硝化反应速率越快：C／N比过高会严重抑制亚硝化反应。     

钢铁企业循环经济影响因素及作用路径研究

基于企业实力、环境管制、公众参与及绿色竞争等理论,提出了钢铁企业循环经济影响因素及作用路径研究模型,并采用偏最小二乘路径分析法,以我国54家钢铁企业2013年数据为例展开实证研究。结果表明:企业特征因素、市场竞争因素、政府因素对钢铁企业循环经济发展水平有直接影响,且影响效应依次递减;公众因素对循环经济发展水平有间接影响,且影响效应较低。同时,验证了此方法的科学性与适用性。     

稳健GM(1,1)模型在工业固废产生量预测中的应用

本文提出了稳健灰色模型,并用该模型建立了山东省工业固体废物产生量数学模型。结果表明,稳健ＧＭ（１,１）模型比通常的ＧＭ（１,１）模型更具预测应用价值。     

基于间隔偏最小二乘法短程硝化反硝化中无机盐氮的近红外光谱

采用序批式活性污泥反应器(SBR)研究短程硝化反硝化系统中稳定条件下无机盐氮的变化规律,利用小波去噪法对无机盐氮的近红外光谱进行预处理,并用间隔偏最小二乘法(iPLS)建立无机盐氮含量的校正模型(iPLS模型).结果表明:小波去噪法对原始光谱中的部分噪声进行滤除,从而提高了模型的精度.采用最小二乘法对预处理后的光谱进行建模,优选出最佳波长区间并将光谱划分为20个子区间,优选出的氨氮特征波数为8243~8663cm-1,亚硝酸盐氮特征波数为4000~4420cm-1.所建模型对氨氮、亚硝酸盐氮校正时的相关系数(rc)分别达到0.9582、0.9544,校正均方根误差(RMSECV)分别为0.0321、0.0406;预测时的相关系数(rp)分别为0.9184、0.8816,预测均方根误差(RMSEP)分别为0.0790、0.0451.采用实际污水为原水时校正模型对氨氮、亚硝酸盐氮预测的相关系数(rp￠)分别为0.9190、0.8739,预测均方根误差(RMSEP′)分别为0.0578、0.0229.模型对氨氮和亚硝酸盐氮的预测总体效果较好.用小波去噪和最小二乘法建立模型不仅能有效减少建模所需变量数、缩短运算时间,而且模型预测精度较高,为无机盐氮的快速测定提供了一种可行的分析技术.     

偏最小二乘(PLS)回归方法在中国东部植被变化归因研究中的应用

植被变化往往受到不同气候变量的综合作用,人类活动影响又使得植被对气候响应变得更为复杂,如何准确判别各种影响因素的相对重要性是植被变化归因研究中的一个关键点.研究基于偏最小二乘(PLS)回归方法,使用1982-2006 年的归一化植被指数(NDVI)数据,分析了降水、气温、日照、相对湿度和风等气候变量对中国东部植被变化的相对影响,并选取了NDVI变化较为典型的区域,量化了农业活动对该地区植被变化的相对贡献.PLS回归方法兼具了主成分分析和多元回归的优点,克服了众多自变量之间存在强烈交互相关导致的多元共线性问题.研究结果表明:① 1982-2006 年间,中国东部逐月NDVI的年际变化呈现出明显的南北差异.在12、1-5 月,NDVI以显著上升为主,上升区域主要位于淮河以北.在6-10月,NDVI以显著下降为主,下降区域主要为淮河以南的部分区域,特别是6 月江苏一带NDVI的大范围下降尤为明显.不过与NDVI发生显著变化的区域相比,更多区域的NDVI在各月并没有出现显著变化.② 在NDVI显著上升的站点,对NDVI变化最具解释意义的气候变量为气温,特别是冬末春初(2-3 月)的升温对黄淮海区域NDVI的显著上升具有主导控制作用.而对于NDVI 显著下降的站点,多数都不能从气候角度解释这些区域的NDVI 变化.③ 江苏省NDVI在6 月出现的大范围显著下降,与农业种植结构的调整,主要是棉花种植面积的减少以及油菜面积的增加具有显著关系.     

Use of linkage mapping and centrality analysis across habitat gradients to conserve connectivity of gray wolf populations in western North America

Centrality metrics evaluate paths between all possible pairwise combinations of sites on a landscape to rank the contribution of each site to facilitating ecological flows across the network of sites. Computational advances now allow application of centrality metrics to landscapes represented as continuous gradients of habitat quality. This avoids the binary classification of landscapes into patch and matrix required by patch-based graph analyses of connectivity. It also avoids the focus on delineating paths between individual pairs of core areas characteristic of most corridor- or linkage-mapping methods of connectivity analysis. Conservation of regional habitat connectivity has the potential to facilitate recovery of the gray wolf (Canis lupus), a species currently recolonizing portions of its historic range in the western United States. We applied 3 contrasting linkage-mapping methods (shortest path, current flow, and minimum-cost-maximum-flow) to spatial data representing wolf habitat to analyze connectivity between wolf populations in central Idaho and Yellowstone National Park (Wyoming). We then applied 3 analogous betweenness centrality metrics to analyze connectivity of wolf habitat throughout the northwestern United States and southwestern Canada to determine where it might be possible to facilitate range expansion and interpopulation dispersal. We developed software to facilitate application of centrality metrics. Shortest-path betweenness centrality identified a minimal network of linkages analogous to those identified by least-cost-path corridor mapping. Current flow and minimum-cost-maximum-flow betweenness centrality identified diffuse networks that included alternative linkages, which will allow greater flexibility in planning. Minimum-cost-maximum-flow betweenness centrality, by integrating both land cost and habitat capacity, allows connectivity to be considered within planning processes that seek to maximize species protection at minimum cost. Centrality analysis is relevant to conservation and landscape genetics at a range of spatial extents, but it may be most broadly applicable within single- and multispecies planning efforts to conserve regional habitat connectivity.     

基于高光谱的土壤全氮含量估测

基于高光谱(350~2 500 nm)数据,研究了我国中、东部地区5种主要类型土壤全氮含量与高光谱反射率之间的定量关系,构建了基于偏最小二乘法(PLS)、BP神经网络(BPNN)和特征光谱指数的土壤全氮含量估算模型。结果表明,以500~900 nm、1 350~1 490 nm区域波段反射率经Norris滤波平滑后的一阶导数光谱为基础,构建的基于PLS和BPNN的土壤全氮含量估算模型精度较高,建模决定系数分别为0.81和0.98;独立观测资料检验结果显示,模型预测决定系数分别为0.81和0.93,均方根误差RMSE为0.219 g·kg^-1和0.149 g·kg^-1,相对分析误差RPD为2.28和3.36,说明PLS和BPNN模型对土壤全氮含量具有较高的预测精度。在光谱指数的分析中,基于近红外872 nm和1 482 nm 两个波段的差值光谱指数DI(NDR₈₇₂,NDR₁₄₈₂)对土壤全氮含量最敏感,建模决定系数、预测决定系数、RMSE和RPD分别为0.66、0.53、0.31 g·kg^-1和1.60。比较而言,三种方法估算土壤氮含量的精度顺序为BPNN模型＞PLS＞DI(NDR₈₇₂,NDR₁₄₈₂),基于PLS和BPNN两种方法建立的土壤全氮含量高光谱估测模型具有较高的精度,可以用来精确估算土壤全氮含量;基于两波段构建的DI(NDR₈₇₂,NDR₁₄₈₂)预测效果低于前两者,但也可以用来粗略估测土壤中的全氮含量。     

An Optimization Method for Estimating Constituent Mean Concentrations in Base Flow‐Dominated Flow

Pollutant coefficients have been widely used to assess runoff nonpoint source pollution from individual land uses (e.g., agricultural, residential) of a watershed. Pollutant coefficients, known as event mean concentrations (EMCs), were developed by the U.S. Environmental Protection Agency's Nationwide Urban Runoff Program (NURP) to serve as a national measure for characterizing pollutant loading in a receiving water body. The term “baseflow pollutant coefficient (BPC)” is used in this study as a surrogate for EMC to describe mean concentration of pollutants in base flow‐dominated flow. A method for characterizing base flow quantity and quality for different land uses was explored using inverse modeling with two optimization techniques (a least square method and a genetic algorithm [GA] optimization), land use information, and streamflow quantity and quality data. The inverse model was formulated as a constrained minimization problem and demonstrated with data for 15 watersheds in Indiana. Results showed that estimated pollutant coefficients are comparable to the published literature. This indicates that the proposed method has the potential to effectively estimate constituent mean concentrations for pollutant load determination in gauged and ungauged watersheds, albeit more analysis with larger and more robust datasets is desirable to further refine and validate the accuracy of the approach.     

基于地统计模型的上海大气污染物多建模方法的比较

地统计模型被广泛应用于环境空气污染物暴露模拟,但不同建模方法及其模拟结果之间的对比研究较少.基于上海2016~2019年55个环境空气监测点位的NO₂和PM_2.5观测数据,以及交通路网、排放源兴趣点和卫星数据等地统计变量,应用偏最小二乘回归（PLS）、监督学习线性回归（SLR）和机器学习随机森林（RF）这3种建模方法创建年暴露模型,并进一步应用普通克里金插值（OK）法分析模型残差,构建复合模型.应用交叉验证对模型的模拟效果进行检验,选取每一种建模方法的最优模型结构（是否应用OK）作为最终模型.结果表明,NO₂模型中表现最好的是RF-OK （R_mse²为0.70~0.82）和PLS-OK模型（R_mse²为0.78~0.84）;PM_2.5模型中PLS模型（R_mse²为0.62~0.71）优于SLR-OK （R_mse²为0.40~0.79）和RF-OK （R_mse²：0.31~0.56）模型.应用3种建模方法对上海1 km网格开展年暴露模拟和对比,NO₂模型间模拟结果的相关性（r为0.82~0.91）高于PM_2.5模拟结果的相关性（r为0.66~0.96）.基于3种模型2019年的模拟结果,评估了上海NO₂和PM_2.5的人群暴露水平.