基于正序阻抗幅值比的风电场送出线路纵联保护

针对风电场送出线路纵联保护在数据延时传输及异常采样数据下保护性能不佳的问题,提出了基于正序阻抗幅值比的纵联保护。通过分析风电系统送出线路发生区内外故障时正序阻抗幅值特征,得出区内外故障下双端正序阻抗幅值差异比特征不同。引入综合层次聚类(balanced iterative reducing and clustering using hierarchies, BIRCH)方法剔除正序阻抗幅值序列中异常采样数据,形成不含异常采样数据的故障时间序列聚类特征,并结合双端正序阻抗幅值差异比特征,构造不受数据延时传输影响的纵联保护判据。仿真结果表明,所提出的纵联保护不受系统运行工况、故障类型、数据延时传输及异常采样数据的影响。在过渡电阻达到150  时,所提出的纵联保护仍能正确判别故障方向,具有较强的抗噪性能,适用于含风电接入的弱馈型电力系统。     

基于重建误差的任务型对话未知意图检测

现有未知意图检测模型通常将语句映射到向量空间,并使用局部异常因子算法定义密度较小的特征点为未知意图,但经交叉熵损失训练的已知意图特征簇更加狭长,簇内的整体间距、密度和分散情况不均匀,进而增加了检测难度。针对上述问题,提出一种基于自动编码器重建误差的未知意图检测模型。在训练阶段,使用融入标签知识的联合损失函数训练已知意图分类器,使已知意图特征类间距离大且类内距离小,并利用这些特征训练一个仅能获取已知意图信息的自动编码器。在测试阶段,利用自动编码器将重建误差较大的样本视为未知意图,其余样本视为已知意图正常分类。在SNIPS数据集上的实验结果表明,在已知意图占比为25%、50%、75%时,该模型的Macro F1得分相比于表现最优的增强语义的高斯混合损失基线模型分别提升了16.93%、1.14%和2.37%,能够检测到更多的未知意图样本,同时在类别分布极不平衡的ATIS数据集上也有较好的性能表现。     

汉维统计机器翻译中的形态学处理

针对汉语和维吾尔语语序差别(前者是主-谓-宾结构,后者是主-宾-谓结构)及形态差别较大的问题,通过编写调序规则将汉语调整为主-宾-谓结构,将维吾尔语单词切分为词干、词缀等更小的词素单元来训练统计模型,同时测试词素的切分粒度对翻译性能的影响.实验结果表明,对汉语句法结构的调整及以词干、词缀等更小的词素形式参与训练可以有效...     

Comparison between Two Spectral Mixture Analysis Methods based on IDL and MATLAB

Interactive Data Language (IDL) is a language in the development of application based on multiplatform and object\|oriented,which has significant advantages in data analysis and visualization.The MATLAB is a software with powerful features in the image of processing and programming in complex numerical analysis,which based on matrix calculation.Programming in combining the IDL with MATLAB,meanwhile using the Extended Linear Mixed Model for endmember unmixing in Hyperion images.To verify the results of endmember unmixing,adopted the Fully Constrained Least Squares for comparative analysis.The results showed that：The method of programming in combining the IDL with MATLAB not noly possess the advantages of both but also enhance the efficiency in programming,it is conducive to remote sensing image processing.Meanwhile endmember unmixing results show that：the Extended Linear Mixing Model unmixing has a higher accuracy when the proportion of local category in the image is large.On the contrary,the Fully Constrained Least Squares unmixing has a higher accuracy.     

面向海量数据的电子政务云平台研究

针对传统电子政务平台所采用的关系型数据库在处理海量数据时存在性能瓶颈问题,利用Hadoop分布式平台在处理海量数据方面的优势,结合HDFS分布式文件系统、Map/Reduce并行计算模型和Hive仓库技术,设计关系型数据库与Hadoop相结合的电子政务云平台,两者协同提供海量数据查询操作和存储服务,从而降低了关系型数据库服务器的负载压力,增强电子政务平台的扩展性。通过实验证明,Hadoop能大大提高电子政务云平台的查询效率。进一步分析该设计方案中影响查询效率的因素,为深入研究基于Hadoop构建高效的电子政务云提供参考。     

药物信息服务实践应用的信度与效度分析

为建立科学的药物信息服务评价体系,我们采用问卷调查法对新疆医科大学第四附属医院药房及患者信息进行系统性调查,利用Excel和SPSSStatistics进行数据整理分析,采用Cronbach＇sAlpha系数、因子分析等方法对问卷做了信度和效度评价分析,初步提出了医院药物信息服务的要素．结果,本次研究的过程和方法为他人在药物信息乃至其它信息分析中提供重要的参考依据．     

异构数据库迁移的设计和实现*

介绍了基于B/S模式异构数据库迁移的具体实现, 阐述了分布式数据库概念、JDBC原理和数据库转换技巧;详细说明了数据转换规则、数据库连接、迁移过程、断点续传。     

Water and Sediment Quality in Lakes along the Middle and Lower Reaches of the Yangtze River, China

Water and sediment samples were collected from 45 lakes along the middle-lower Yangtze River, China. Each lake was sampled seasonally, over a period of one year. Water quality variables and metals in sediments were measured. Lakes along the lowest part of the river, within the Yangtze River Delta, had highest nutrient concentrations and were eutrophic or hypereutrophic. Lakes displayed a gradient in many water chemistry variables, from the middle to the lower Yangtze River. Lakes of the Delta region had the highest conductivity, sulfate, turbidity, and Chl-a values, and the lowest dissolved oxygen concentrations. A number of lakes near urban areas in the study region also displayed similar conditions. Lakes polluted by heavy metals were found in the upper part of the lower Yangtze River and had high Cu, Cr, and Co concentrations in sediments. The mean I_geo (index of geoaccumulation) values for Cu, Cr, and Co classes ranged from 0 to 4, indicating moderate to heavy contamination, contributed mainly from untreated industrial waste water produced within the lake catchments. Lakes of the middle Yangtze River are generally in relatively better condition, except for those around urban zones, which experience higher nutrient and heavy metal loading. The spatial distribution of lake conditions in the area can be related to the policy of regional economic development. The Delta region in China is developed and includes such cities as Shanghai and Suzhou. Heavy industries have moved into the interior region of China too, and development of modern cities is now occurring under some level of environmental protection. Nevertheless, lakes in the central part are becoming seriously polluted with both heavy metals and nutrients because local authorities promote industrialization and urbanization to improve economic conditions, while often ignoring environmental protection. Pollution is increasingly occurring in upstream reaches, a tendency that will bring more environmental problems. Interior lakes of China require immediate attention to prevent further declines in water quality.     

塔里木盆地塔中北坡志留系致密碎屑岩地震反演技术及应用

塔里木盆地塔中北坡已经发现了致密-低孔低渗岩性油气藏,主要层位为志留系下统柯坪塔格组下砂岩段,本区岩性油气藏具有典型的砂泥岩薄互层特征,而且横向变化大、非均质性极强,含油砂岩的声波阻抗和泥岩的声波阻抗存在较大范围的重叠,难以区分。针对本区岩性圈闭,致密碎屑岩储层的地质、地球物理特点存在的难点,开展地震储层综合预测研究,提出将基于部分叠加数据的叠前同时反演和地质统计学结合起来的致密碎屑岩储层地震反演技术,通过实际应用,验证了该反演技术对储层预测的有效性。     

高密度采集噪音特征分析与压制技术研究

高密度采集通过加大空间域、时间域的数据采集密度,很好地记录了资料的噪音特性,便于对其进行压制处理,但这同时也增加了资料的数据量,加大了室内处理难度。对此,本文将根据资料的噪音特性对其进行组合压制,这样既能很大程度地提高资料的信噪比,改善资料品质;同时又能减少数据量,便于室内处理。