Spatial analysis of oil reservoirs using detrended fluctuation analysis of geophysical data (original) (raw)
Related papers
Spatial analysis of oil reservoirs using DFA of geophysical data
We employ Detrended Fluctuation Analysis (DFA) technique to investigate spatial properties of an oil reservoir. This reservoir is situated at Bacia de Namorados, RJ, Brazil. The data corresponds to well logs of the following geophysical quantities: sonic, gamma ray, density, porosity and electrical resistivity, measured in 56 wells. We tested the hypothesis of constructing spatial models using data from fluctuation analysis over well logs. To verify this hypothesis we compare the matrix of distances among well logs with the differences among DFA-exponents of geophysical quantities using spatial correlation function and Mantel test. Our data analysis suggests that sonic profile is a good candidate to represent spatial structures. Then, we apply the clustering analysis technique to the sonic profile to identify these spatial patterns. In addition we use the Mantel test to search for correlation among DFA-exponents of geophysical quantities.
The spatio-temporal variations in geophysical well-log signals, which often reflect their scale invariant properties, can be well studied with multifractal analysis. In this study, we have carried out fractal and multifractal studies using detrended fluctuation analysis (DFA) and multifractal DFA (MFDFA) respectively. While the DFA primarily facilitates to understand the intrinsic self-similarities in non-stationary signals like well-logs by determining the fractal scaling exponents in a modified least-squares sense, the MFDFA, which in fact, is a generalization of DFA, provides a comprehensive understanding of the multifractal behaviour of the signals through multifractal singularity spectrum as well as the Hurst exponents. DFA and MFDFA have been applied to gamma-ray log and neutron porosity logs of two wells (well B and well C), located in the western offshore basin, India, to study the nature of the subsurface formation properties, vis-à-vis their multifractal behaviour. The estimated DFA fractal scaling exponents, represented in the form of contour plots enable easy identification of the depths to the tops of reservoir zones. On the other hand, the multifractal singularity spectra provide a unique platform for an improved interpretation of logs in terms of their sedimentation pattern and lithological differences. This has been tested with gamma-ray log data of wells B and C. We show that the multifractal behaviour of gamma-ray log is largely influenced by the presence of shale and variations in the subsurface sedimentation pattern. Similarly, the role of gas in a pay zone on the multifractal behaviour was established by comparing the multifractal singularity spectra of the original neutron porosity log and a synthetic neutron log (which we call gas-corrected log), generated using density log. The MFDFA of only that portion of the original neutron log representing the pay zone and its gas-corrected equivalent unequivocally suggest that the presence of gas in the reservoir zones weakens the multifractal behaviour of neutron porosity logs. This emphasizes the significance of multifractal studies of well-logs for effective reservoir characterization. The observed multifractal behaviour in all logs is found to be due to the presence of long-range correlations in the data.
Revealing spatial variability structures of geostatistical functional data via Dynamic Clustering
2011
In several environmental applications data are functions of time, essentially con- tinuous, observed and recorded discretely, and spatially correlated. Most of the methods for analyzing such data are extensions of spatial statistical tools which deal with spatially dependent functional data. In such framework, this paper introduces a new clustering method. The main features are that it finds groups of functions that are similar to each other in terms of their spatial functional variability and that it locates a set of centers which summarize the spatial functional variability of each cluster. The method optimizes, through an iterative algorithm, a best fit criterion between the partition of the curves and the representative element of the clusters, assumed to be a variogram function. The performance of the proposed clustering method was evaluated by studying the results obtained through the application on simulated and real datasets.
Geophysical Journal International, 2002
The spatial cross-correlation and power spectra of porosity and log(permeability) sequences are analysed for a total of 750 m of reservoir rock drill-core from four vertical wells in the Brae Formation, an important coarse-grained clastic North Sea hydrocarbon reservoir rock. The well core sequences are 80t4 per cent cross-correlated at zero lag and have power-law-scaling spatial power spectra S(k)31/k b , b#1t0.4, for spatial frequencies 5 km x1 <k<3000 km x1 . The strong spatial cross-correlation of porosity and log(permeability) and the systematic power-law scaling of log(permeability) spatial fluctuation spectra fit into a broad physical context of (1) the 1 / k spectral scaling observed in several hundred well logs of sedimentary and crystalline rock recorded world-wide; (2) the 1/ f spectral scaling of temporal sequences in a wide range of physical systems; and (3) analogy with power-law-scaling spatial fluctuation spectra in a wide range of critical-state thermodynamic systems. In this physical context, the spatial fluctuations of log(permeability) of clastic reservoir rock are interpreted as due to longrange correlated random fracture-permeability networks in a fluid-saturated granular medium where the range j of spatial correlation is effectively infinite.
Nafta-Gaz, 2016
Determination of the statistical similarity of the physicochemical measurement data of shale formations based on the methods of cluster analysis The paper presents the application of the methods of statistical analysis to the determination of the similarity of measurement data, using boreholes providing access to shale formations as an example. The proposed methodology is based on two statistical techniques: the factor analysis and the cluster analysis. The first method allows the reduction of the number of measurements variables in order to eliminate the redundancy of the data. The second one allows grouping the wells on the grounds of factor variables defining the similarity features of the analysed wells. The available results of geochemical measurements for nine wells providing access to shale structures have been used as measurement data.
Reservoir characterization based on seismic spectral variations
The seismic frequency spectrum provides a useful source of information for reservoir characterization. For a seismic profile presented in the time-space domain, a vector of the frequency spectrum can be generated at every sampling point. Because the spectrum vectors at different time-space locations have different variation features, I attempt for the first time to exploit the variation pattern of the frequency spectrum for reservoir characterization, and test this innovative technology in prediction of coalbed methane (CBM) reservoirs. The prediction process implicitly takes account of the CBM reservoir factors (such as viscosity, elasticity, cleat system, wave interference within a coal seam, etc.) that affect the frequency spectrum, but strong amplitudes in seismic reflections do not necessarily show any influence in clustering analysis of spectral variation patterns. By calibrating these variation patterns quantitatively with CBM productions in well locations, we are able to characterize the spatial distribution of potential reservoirs.
A spatially focused clustering methodology for mining seismicity
Engineering Geology, 2018
Mining seismicity is routinely observed to cluster in space and time due to the spatially distinct rock mass failure processes associated with the temporally dependent process of mining. Assessment of clustered seismicity is important to develop an understanding of and to quantify seismic hazard that is associated with mining. This article presents a density-based clustering method that is applicable to the assessment of 3D spatial distributions of short-term seismicity. The methodology presented in this article is developed from existing approaches that address the general limitations of density-based clustering algorithms. Synthetically generated seismicity allows for the assessment of the methodology with respect to external and internal performance measures. The clustering of a dataset with known attributes allows for confidence to be developed in the capability of the clustering method. Additionally, this internal performance evaluation can represent the relative accuracy of outcomes without prior information concerning dataset attributes. The clustering method is applied to two case studies of mining seismicity. These cases illustrate the general applicability of the clustering method along with the value of evaluating internal performance measures when optimising the selection of parameters and understanding the sensitivity of clustering outcomes to these choices. which uses a simple metric to define a space-time distance between
Unconventional Resources Technology Conference (URTeC)
Tying detailed well log measurements to lower resolution but a really extensive 3D seismic data volumes is key to quantitative seismic interpretation. Ties using a poststack or prestack convolution model are routine, while supervised classification tying well data to seismic attributes using neural networks and geostatistics are also well established. However, unsupervised classification ties where the objective is to identify unknown patterns in the data is less well established. In this paper, we use an automatic learning Gaussian Mixture Model to statistically characterize the well logs, evaluate the probability distribution functions of different lithologies and then tie them to corresponding 3D seismic attribute volumes. We precondition our four-dimensional data by projecting onto two dimensions using Independent Component Analysis.