基于改进多分类孪生支持向量机的测井岩性识别方法研究与应用

岩性识别是认识地层及求解储层参数的基础,受地质环境复杂性和非均质性影响,测井曲线间存在着大量的信息冗余,数据集类间分布不平衡,常用的分类算法无法满足实际需求.针对常用分类算法容错性差,识别岩性单一和无法有效解决类间不平衡的问题,本文改进合成少数过采样技术(Synthetic Minority Over Sampling Technique,SMOTE)来处理数据集,可得到类间平衡的新数据集,并提出一种新的模糊隶属度函数改进模糊孪生支持向量机,在北美Hugoton油气田实际测井数据的基础上,用改进多分类孪生支持向量(Improve Multi Class Twin Support Vector Machine,IMCTSVM)综合自然伽马(GR)、电阻率(RL)、光电效应(PE)、中子密度孔隙度差异(DPHI)和平均中子密度孔隙度(PHIND)五种测井参数,以及相对位置(RELPOS)和非海洋/海洋指标(NM_M)两种地质约束变量,识别出9种岩性.将识别结果与传统支持向量机、深度神经网络等方法进行对比与分析,发现IMCTSVM算法优于上述两种分类算法,取得了较好的识别效果.     

Reservoir characterization of a CO2 storage aquifer: The Upper Triassic Stuttgart Formation in the Northeast German Basin

Ketzin, in the Northeast German Basin (NEGB), is the site for pilot injection of CO₂ (CO₂SINK project) into a saline aquifer (the Upper Triassic Stuttgart Formation) situated at a depth of about 630–700 m. This paper reports the baseline characterization of the reservoir formation based on new core material and well-logs obtained from one injection well and two observations wells, drilled at a distance from 50 m to 100 m from each other. The reservoir is lithologically heterogeneous and made up by fluvial sandstones and siltstones interbedded with mudstones showing remarkable differences in porosity. The thickest sandstone units are associated with channel sandstone, whose thickness varies over short lateral distances. In-depth petrographic, mineralogical, mineral-chemical, and whole-rock geochemical analysis were performed focusing on the sandstone intervals, which display the best reservoir properties for CO₂ injection. The dominantly fine-grained and well to moderately-well sorted, immature sandstones classify as feldspathic litharenites and lithic arkoses. Quartz (22–43 wt.%), plagioclase (19–32 wt.%), and K-feldspar (5–13 wt.%) predominate mineralogically. Muscovite plus illite and mixed-layer minerals are omnipresent (4–13 wt.%). Quartz, feldspar, as well as meta-sedimentary and volcanic rock fragments comprise the most abundant detrital components, which often are rimmed by thin, early diagenetic coatings of ferric oxides, and locally of clay minerals. Feldspar grains may be unaltered and optically clear, partially to completely dissolved, partially altered to sheet silicates (mainly illite), or albitized. Analcime and anhydrite constitute the most widespread, often spatially associated pore-filling cement minerals. Authigenic dolomite, barite, and coelestine is minor. The percentage of cements ranges in total from about 5 vol.% to 32 vol.%. Except of samples intensely cemented by anhydrite and analcime, total porosities of the sandstones range from 13% to 26%. The fraction of intergranular porosity varies between 12% and 21%. About 1–5% porosity has been generated by dissolution of detrital plagioclase, K-feldspar, and volcanic rock fragments. The comparatively large modal abundance of feldspars, micas, chlorite, clay minerals, Fe–Ti-oxides, and analcime account for the richness in Ti, Al, Fe, Mg, Na, and K, and the paucity in Si, of the Stuttgart sandstones relative to mature sandstones. Altogether, these sandstones are comparatively rich in minerals that may potentially react with the injected CO₂.     

基于随机森林算法的陆相沉积烃源岩定量地震刻画:以东海盆地长江坳陷为例

烃源岩的定量地震刻画对于勘探开发区块的优选、盆地油气资源量的估算都具有重要意义.陆相沉积环境下的浅湖或半深湖相的烃源岩横向变化快,其空间展布需要依靠钻井约束下的反射地震进行刻画,但是其地震弹性特征与岩性和有机质含量的映射关系呈现高度非线性化,因而很难利用传统基于地震岩石物理模型驱动的烃源岩地震预测方法进行有效刻画.本文以低勘探区的东海盆地长江坳陷为例,提出了一种在数据驱动的机器学习框架下,综合利用地质约束、钻井录井、测井、地球化学和叠前地震数据进行烃源岩的定量地震刻画的工作流程.其核心思想是利用随机森林集成学习算法对小样本数据表现优异的特征,以井位处的测井弹性数据(纵波速度和密度)、岩性、地球化学标定的总有机碳含量(TOC)为样本标签数据,在地质导向约束下通过随机森林算法生成学习网络,并将该网络与叠前地震反演结果相结合,采取先预测泥岩再预测总有机碳含量的“两步走”策略,完成对烃源岩空间分布及其非均质性的定量地震刻画,并对预测结果的不确定性进行评价.测试结果显示,随机森林算法相较于其他的机器学习算法能够更准确的识别陆相沉积地层的泥岩,并比传统的利用阻抗转化方法获得更可靠的总有机碳含量预测结果.     

Influence of borehole quantity and distribution on lithology field simulation

This paper aims to study the influence of the number and distribution of drill samples on the simulation accuracy of the lithology field. This research mainly applies the variation function method in geo-statistics, and determines important indicators such as the variation, and then the lithology field is simulated by sequence index simulation. It is shown that (1) simulation error decreases with the increase of sampling density; (2) at the scale and complexity of this study, when the sampling density reaches 40 /km2, the average error of the lithology field simulation can be less than 2.0%; (3) in the study mode of examples, the simulation results of random sampling in the whole region are the most ideal, with an average error of 5.4%. The average error of the simulation results of the centralized sampling is about 10 times that of the random sampling method; (4) known from the influence analysis of the degree of study sample unevenness influence on the imitation results, under the same sample size, the simulation error decreases with the increase of the most adjacent index. When the nearest index reaches 1, the simulation error will be less than 6%, and the error variable range is within 3%.     

Détection des transitions lithologiques par l'analyse de la composante fractale des diagraphies par transformée continue en ondelettes

The frequency analysis of many log data permits to verify that their stochastic component show ‘power-law-type’ spectral densities, characteristic of $1/f$ noise. They can be modelled by fractional Brownian motions. Continuous Wavelet Transformation (CWT) provides us with very efficient methods to determine the local spectral exponents of these scaling laws. These new attributes are related to the local fractality of these signals. We first present some theoretical results and an application to a fractional Brownian motion. The second application concerns a dataset recorded in the MAR203 borehole. We show that clustering of these new pseudo-logs leads to a good resolution between different lithofacies. To cite this article: N. Zaourar et al., C. R. Geoscience 338 (2006).     

利用岩性阻抗识别砂岩、泥岩以及砂体叠置——以北部湾盆地涠洲A油田为例

针对涠洲A油田储层平面非均质性强、存在砂体叠置,以及单纯的波阻抗反演不能十分有效地识别砂泥岩和叠置砂体,提出了将纵波阻抗与横波阻抗通过一定的数学变换获得岩性阻抗的方法,证实了该岩性阻抗反演结果与实钻井吻合程度高,有效区分了砂岩和泥岩,并成功地预测了涠洲A油田砂体叠置的关系。     

Lithological interpretation of crustal composition in the Fennoscandian Shield with seismic velocity data

In this study, we report the results of an investigation of lithological interpretation of the crust in the central Fennoscandian Shield (in Finland) using seismic wide-angle velocity models and laboratory measurements on P- and S-wave velocities of different rock types. The velocities adopted from wide-angle velocity models were compared with laboratory velocities of different rock types corrected for the crustal PT conditions in the study area. The wide-angle velocity models indicate that the P-wave velocity does not only increase step-wise at boundaries of major crustal layers, but there is also gradual increase of velocity within the layers. On the other hand, the laboratory measurements of velocities indicate that no single rock type is able to provide the gradual downward increasing trends. Thus, there must be gradual vertical changes in rock composition. The downward increase of velocities indicates that the composition of the crust becomes gradually more mafic with increasing depth. We have calculated vertical velocity profiles for a range of possible crustal lithological compositions. The Finnish crustal velocity profiles require a more mafic composition than an average global continental model would suggest. For instance, on the SVEKA'81 transect, the calculated models suggest that the crustal velocity profiles can be simulated with rock type mixtures where the upper crust consists of felsic gneisses and granitic–granodioritic rocks with a minor contribution of amphibolite and diabase. In the middle crust, the amphibolite proportion increases. The lower crust consists of tonalitic gneiss, mafic garnet granulite, hornblendite, pyroxenite and minor mafic eclogite. Assuming that these rock types are present in sufficiently extensive and thick layers, they would also have sufficiently high acoustic reflection coefficients for generating the generally well-developed reflectivity in the crust in the central part of the shield. Density profiles calculated from the lithological models suggest that there is practically no density contrast at Moho in areas of the high-velocity lower crust. Comparison of reflectors from FIRE-1 and FIRE-3 transects and the velocity model from SVEKA'81 wide-angle transect indicated that the reflectors correlate with velocity layering, but the three-dimensional structures of the crust complicate such comparisons.     

Holocene Lake Level Fluctuations Recorded in the Sediment Compositionof Lake Juusa,Southeastern Estonia

The main aim of this study was to reconstruct the history of water level fluctuations of a small dimictic lake during the Holocene based on the lithological and geochemical composition of sediments. We studied lithological and geochemical parameters in sediment cores extracted along a transect through Lake Juusa in southeastern Estonia. Analysis revealed that LOI₅₅₀ and CaCO₃ content in the accumulated sediment are dependent on water depth. At depths greater than 3 m, the relationship between organic carbon and CaCO₃ is inverse, whereas no relationship exists at shallower depths. Lake-level changes are well depicted on the grain-size distribution profiles and macrofossil composition. An increase in the >‰315μm fraction appeared during the initial phases of water level fluctuations, followed by an increase in the 100–315μm fraction. The >‰36μm fraction dominated the silt fraction. Higher OC/N ratios in shallow water sediments reflect the dominance of macrophytes vegetation.     

广西金牙金矿床形成条件浅析

文章分析了与金牙金矿矿床形成有关的岩性、构造和岩浆岩;探讨了金的来源、活化条件和迁移形式;利用地质地球化学热力学计算了成矿的物理化学参数。     

Geologic influences on fluvial hydrology and bedload transport in small mountainous watersheds,Northern New Mexico,U.S.A.

Discharge characteristics in six adjacent mountainous watersheds in northern New Mexico, U.S.A., vary substantially between basins underlain by different lithologies. Relatively resistant gneisses and granites underlie two basins (drainage areas: 43 and 94 km²) that have high unit discharge (0·010 to 0·14 m³s^?1 km^?2), high bankfull discharge, and sustained high discharge. Less resistant sandstones and shales underlie four basins (drainage areas: 96 to 215 km²) that have relatively low unit discharge (0·001 to 0·005 m³s^?1 km^?2), relatively low bankfull discharge, and peak discharges that are not sustained as long as those in the crystalline terrane. Analysis of snowmelt-runoff water budgets suggests that three factors control hydrologic conditions in the basins. First, area-elevation distributions appear to control the timing and amounts of water input. These distributions probably reflect the erosional resistance of the different lithologies. Second, lithology appears to control runoff production in areas having minor amounts of storage. Third, glacial deposits in headwater regions control discharge duration and timing via storage and return flow releases. The amount of return flow released by glacial deposits, however, is probably controlled by the permeability of underlying bedrock. Therefore it appears that the duration, timing, and magnitude of discharge events in the study area are controlled both directly and indirectly by lithology. Stream power and shear stress estimates derived from bankfull discharge and bed-material size data suggest that higher bedload transport rates and larger bedload particle sizes exist in streams draining crystalline rocks than in streams draining sedimentary terrane. It appears that source-area lithology, by controlling discharge production, also influences stream power, bedload transport capabilities, and therefore total amounts of bedload transport.