磨溪龙王庙组碳酸盐岩储层多尺度离散裂缝建模技术及其应用

在以大斜度井和水平井为主要开发井型的缝洞型碳酸盐岩气藏中,要想获取裂缝在井点不同空间位置的产状较困难,裂缝精细描述存在不准确性,这些均影响了对气藏渗流通道的刻画,制约了边水气藏的科学、均衡开发。以磨溪龙王庙组碳酸盐岩储层为例,利用岩心照片、FMI成像测井、叠前地震各向异性裂缝预测和不连续性检测以及动态监测等资料,在大斜度井、水平井裂缝定性识别的基础上,定量表征了裂缝产状、开度、密度、孔隙度等参数,再结合所获取裂缝参数建立多尺度非结构化网格离散裂缝模型,明确了气藏高、低渗区域分布,优势水侵通道和水侵方式。结果表明：在磨溪龙王庙组气藏离散裂缝模型中,大尺度和中小尺度裂缝均较发育,高渗区呈连片状广泛分布;发育于气藏外围4个方向的9条边水水侵的高渗通道,表现为沿裂缝水窜型和沿溶蚀孔洞均匀推进型2种水侵方式。该方法及研究结果对同类特大型超压有水深层碳酸盐岩气藏裂缝精细描述、水侵优势通道刻画和水侵模式建立等理论和技术研究均具有借鉴意义。     

Energy performance of building envelopes in different climate zones in China

Energy performance of office building envelope designs in the five major climate zones – severe cold, cold, hot summer and cold winter, mild, and hot summer and warm winter – in China was investigated. A major city within each climate zone was selected. These were Harbin, Beijing, Shanghai, Kunming and Hong Kong. Generic building envelopes were developed based on data gathered from building surveys, local energy codes and the ASHRAE Standard. The overall thermal transfer value (OTTV) method and the heating degree-days technique were adopted in the analysis. Cooling and heating requirements due to heat gain/loss through the building envelopes were determined based on the respective OTTV parameters and building load coefficients. Different shape coefficients were also considered. For a typical floor, chiller load due to heat gain through the building envelope varied from 1.0 kW h/m² in Kunming to 23.5 kW h/m² in Beijing, and the heating load ranged from 2.7 kW h/m² in Hong Kong to 124.3 kW h/m² in Harbin.     

东濮凹陷构造特征及古生界有利勘探区带评价

利用地震、钻井等资料,对东濮凹陷进行构造解析,并结合煤系源岩高热演化区的生气强度进行了研究区古生界有利勘探区带预测。研究结果表明：①东濮凹陷具有印支—燕山期挤压分区、古近纪拉张成带、沙四—沙三段沉积期断陷控洼、沙二段沉积期走滑成块的演化特征,从而导致古近系具有多断、多凸、多洼的结构特征。②东濮凹陷在多期构造应力叠合作用下形成“棋盘式”含油气系统,各含油气系统内油气普遍具有“上油下气、早油晚气、连续成藏”的特征,总体上发育4类油气藏,建立了弱改造型盆地“近源聚集、连续成藏、断超控富”的油气富集模式。③东濮凹陷中区构造改造弱,油气富集,南北向挤压断层控制着古生界残余地层的分布,是深层煤层气原生油气藏有利区带;在高热演化区以烃源岩为中心,利用多成藏要素叠合可优选出卫城—文留构造带为有利勘探区。     

一种提高神经网络集成差异性的学习方法

集成学习已经成为机器学习的研究方向之一,它可以显著地提高分类器的泛化性能.本文分析了Bagging及AdaBoost集成方法,指出了这两种方法的缺陷;然后提出了一种新的基于神经网络的分类器集成方法DBNNE,该方法通过生成差异数据增加集成的差异性;另外,当生成一个分类器后,采用了测试方法确保分类器集成的正确率;最后针对十个标准数据集进行了实验研究,结果表明集成算法DBNNE在小规模数据集上优于Bagging及AdaBoost集成方法,而在较大数据集上也不逊色于这两种集成方法.     

A method for evaluating paleo hydrocarbon pools and predicting secondary reservoirs: a case study of the Sangonghe Formation in the Mosuowan area, Junggar Basin

Taking the Jurassic Sangonghe Formation in the Mosuowan-Mobei area of the Junggar Basin as an example, this paper provides a method that evaluates paleo hydrocarbon pools and predicts secondary reservoirs. Through Quantitative Grain Fluorescence (QGF) experiments, well-tie seismic correlation, and paleo structure analysis, the scale and distribution of paleo hydrocarbon pools in the study area are outlined. Combining current structural features and fault characteristics, the re-migration pathways of paleo oil and gas are depicted. Based on barrier conditions on the oil re-migration pathways and current reservoir distribution, we recognize three types of secondary reservoirs. By analyzing structural evolution and sand body-fault distribution, the major control factors of secondary reservoirs are specified and, consequently, favorable zones for secondary reservoirs are predicted. The results are mainly as follows. (1) In the primary accumulation period in the Cretaceous, paleo hydrocarbon pools were formed in the Sangonghe Formation of the Mosuowan uplift and their size and distribution were extensive and the exploration potential for secondary reservoirs should not be ignored. Besides, paleo reservoirs were also formed in the Mobei uplift, but just small scale. (2) In the adjustment period in the Neogene, traps were reshaped or destroyed and so were the paleo reservoirs, resulting in oil release. The released oil migrated linearly northward along the structural highs of the Mobei uplift and the Qianshao low-relief uplift and then formed secondary reservoirs when it met new traps. In this process, a structural ridge cooperated with sand bodies and faults, applying unobstructed pathways for oil and gas re-migration. (3) The secondary hydrocarbon pools are classified into three types: low-relief anticlinal type, lithologic pinch-out type and fault block type. The distribution of the first type is controlled by a residual low uplift in the north flank of the paleo-anticline. The second type is distributed in the lithologic pinch-out zones on the periphery of the inherited paleo uplift. The third type is controlled by fault zones of which the strikes are perpendicular to the hydrocarbon re-migration pathways. (4) Four favorable zones for secondary reservoirs are predicted: the low-relief structural zone of the north flank of the Mosuowan paleo-anticline, the fault barrier zone on the western flank of the Mobei uplift, the Qianshao low-relief uplift and the north area of the Mobei uplift that parallels the fault zone. The study above effectively supports the exploration of the Qianshao low-relief uplift, with commercial oil discovered in the Qianshao1 well. Besides, the research process in this paper can also be applied to other basins to explore for secondary reservoirs.     

支持向量机推广能力估计方法比较

支持向量机是一种新的机器学习算法，与其它学习算法相比，它的最大优点是基于结构风险最小化原则，因而能够保证推广能力。推广能力估计是机器学习中的一个重要问题，是实现自适应调整、参数选择、模型选择的等方法的基础。本文详细比较当前较有影响的几种推广能力估计方法，指出了这些方法适应范围和优缺点，并结合各种方法的原理讨论了推广能力估计可能的发展方向。     

神经网络中将任务学习与模型学习相结合的E-H和E-H-W学习方法

本文从获取好的神经网络泛化能力出发,首先提出了将Hebbian学习与增加问题复杂性统一起来的思想,并通过在总的误差函数中增加一限制函数来实现Hebbian学习.基于此,提出了一种将误差驱动的任务学习与Hebbian规则的模型学习相结合的E-H方法.然后,根据模型学习应同时考虑减小网络复杂性和增加问题复杂性的思想,又提出了一种将误差驱动的学习与Hebbian规则、简单的权退化法结合起来,共同来提高神经网络的泛化能力的E-H-W方法.最后通过大量实例仿真将它们与纯误差驱动的方法、权退化法、其它文献中的相关方法进行了比较.结果表明我们的方法具有最好的泛化能力,是很有效的神经网络学习方法.     

异构分类器堆叠泛化及其在恶意评论检测中的应用

恶意评论检测是预防社会媒体平台给用户带来负面影响的一项重要工作,是自然语言处理的重要领域之一.为解决单分类器实现恶意评论检测时模型精度不稳定、boosting集成模型精度较低的问题,提出一种异构分类器堆叠泛化的方法.该方法用深度循环神经网络将多标签的恶意评论分类问题转变为二类分类,防止了模型精度不稳定;用堆叠泛化集成时单个分类器GRU（Gated Recurrent Unit）和NB-SVM（Naïve Bayes-Support Vector Machine）在模型结构和分类偏差上的差异性,改善了模型精度.在维基百科恶意评论数据集上的对比实验证明：提出的方法优于boosting集成,说明堆叠泛化异构分类器实现恶意评论检测是可行且有效的.     

汽车行业冲压厂的装箱料架规划与配置

根据冲压件的装箱方式及各车型零件间的相似性,对装箱料架进行分类,将其系列化、标准化。分析整车销量与冲压单批次产量之间的关系,总结出每年用于冲压周转所需的装箱料架总数的理论公式,对比总装箱料架的理论值和实际值,了解实际装箱料架数量是否已满足生产要求。根据装箱料架总数的变化趋势,调整冲压生产的频次,按此方式执行,消除装箱料架的闲置现象,大大提升了库房的有效利用率。