Hart, Nicole (Premier Oilfield Group) | Dix, Michael (Premier Oilfield Group) | Mainali, Pukar (Premier Oilfield Group) | Rowe, Harold (Premier Oilfield Group) | Morrell, Austin (Premier Oilfield Group) | Matheny, Mei (Premier Oilfield Group)
The Powder River Basin (PRB) is an asymmetric foreland basin that historically hosted conventional oil plays, however, due to the development of evaluation and completion techniques for unconventional reservoirs, the PRB is being reevaluated for additional reserves. While PRB unconventional development has been limited thus far, the stacked potential of the Cretaceous section has become an intriguing prospect as the oil market recovers. Cretaceous unconventional targets in the PRB include two main source-rock intervals, the Mowry Shale and the Niobrara Formation, as well as tight and shaley siliciclastic reservoirs such as the Frontier, Turner, Shannon, and Sussex sandstones. Understanding mineral variation in these reservoir lithologies is essential for effective formation evaluation. Knowledge of clay and carbonate mineral abundances in the Mowry and Niobrara formations is particularly important, as these exert a first-order control on brittleness and permeability in these source rocks. Elemental data from cuttings have increasingly been used to model mineralogy, brittleness, and organic content in mudstone-dominated sequences. While this approach is well-established in other basins, it has not been rigorously attempted in the PRB.
An X-ray fluorescence (XRF) spectrometer with a customized reference-based calibration was used to collect quantitative data for 29 elements from wellbore cuttings. Mineralogy was then estimated from these elemental data using a stepwise element-parceling logic that was cross-checked against measured X-ray diffraction (XRD) mineralogy results. Elemental proxies for organic richness are also compared to TOC values. These established relationships can be utilized to develop a model for predicting TOC in the Mowry Shale and the Niobrara Formation.
Both XRF mineral modeling and XRD results indicated a wide range of quartz, calcite, clays, plagioclase-dominated feldspar, relatively low dolomite, and significant apatite. The total clay abundance of individual samples can exceed 50 weight percent, with mixed-layer illite/smectite, illite, and lesser amounts of chlorite and kaolinite represented. These preliminary results showed a strong correlation between the XRF-calculated mineralogy from the model and XRD-measured mineralogy in multiple stratigraphic units. The predominance of plagioclase over K-feldspar partially facilitates the accuracy of the model by simplifying the parceling of K2O into illite/smectite and illite. Additional preliminary results indicated that copper, molybdenum, uranium, and nickel correlated best with TOC and these relationships could potentially be used to model TOC in the Mowry and Niobrara.
Understanding mineral variation in these reservoir lithologies is essential because the abundances of clay and carbonate minerals largely dictate brittleness and permeability, and therefore the ability to fracture and produce hydrocarbons from these units. Additionally, a better understanding of the vertical distribution of TOC allows for improved well placement through the targeting of organic rich intervals. This workflow allows for effective formation evaluation through the rapid and economical collection of a versatile data set and model results that can be extrapolated to other wells to better understand the lateral and vertical variations in reservoir mineralogy and TOC across the southern PRB.
Over the years, low resistivity pay has become recognized as a worldwide phenomenon, unfortunately only a few of these reservoirs are successfully identified and evaluated using standard logging data. This is especially true in the Cretaceous clastic reservoirs of the Orinoco Oil Belt. The primary goals of this investigation were to understand the causes of low resistivity in the pay zones, the nature of the gamma ray response, and to investigate the potential of whole-rock elemental data for characterizing these reservoirs.
It is well known that improvements in missed pay identification must include integration of geological, petrophysical, and reservoir engineering data. The first step was to characterize the composition of the Creataceous formation in well-constrained core samples from an older vertical well, X-23. Thin section and X-ray diffraction (XRD) analyses showed the samples to contain significant amounts of detrital clay, most of which is kaolinite. Some authigenic kaolinite may also be present, but is minor. The source of low resistivity in the X-23 reservoir section can be logically interpreted to be formation water present in microporosity associated with the bundant clay minerals.
For the second step, a test was performed on cuttings samples from the Cretaceous reservoir in the recently-drilled X-272 horizontal well. No core was available from this well. Whole-rock elemental data was obtained for 50 elements from 29 washed cuttings samples, using combined WD-XRF (Wavelength-Dispersive X-ray Fluorescence spectroscopy) and ICP-MS (Inductively Coupled Plasma - Mass Spectrometry). Three of these samples were also analyzed for mineralogy by XRD. Results of the combined analyses indicated the samples contained little clay (about 2%), virtually no plagioclase, significant K-feldspar (5-11%), small amounts of carbonate (2-8%), and moderate amounts of heavy minerals. The heavy minerals, as inferred from elemental data (TiO2, Zr, Nb, Th, U and rare earth element (REE)), are likely to be ilmenite, rutile, titanite, zircon, and apatite; monazite, xenotime, garnet, micas, and Th-oxides may be present as well. The low Al2O3 and XRD clay values preclude these elements being primarily associated with clays.
The third step was to assess the gamma and resistivity responses of the logs. In the X-272 horizontal well, the gamma and resistivity responses could not be adequately explained by the composition inferred by the combined mineralogical and elemental analyses. It is therefore suspected that there was significant detrital clay present in the original oil-saturated cuttings samples, but almost all was removed during cleaning of cuttings at wellsite, leaving the sand fraction as a residual sample.
Despite the difficulties in obtaining representative samples from washed cuttings, the elemental data from the sand fraction of the X-272 samples alone shows potential for the definition and chemostratigraphic correlation of distinctive stratigraphic units for the purpose of wellbore positioning. In addition, the utility of the elemental data has provided motivation to develop an improved method of sample cleaning for oil-sands cuttings at wellsite.
MacDonald, Robin (Saudi Aramco) | Hardman, Douglas (Saudi Aramco) | Sprague, Ronald (Saudi Aramco) | Meridji, Yacine (Saudi Aramco) | Mudjiono, Witjaksono (Saudi Aramco) | Galford, James (Halliburton) | Rourke, Marvin (Halliburton) | Dix, Michael (Chemostrat) | Kelton, Michael (Core Laboratories)
Chen, Dingding (Halliburton Energy Services Group) | Hamid, Syed (Halliburton Energy Services Group) | Dix, Michael (Halliburton Sperry Drilling Services) | Quirein, John Andrew (Halliburton Energy Services Group) | Jacobson, Larry A. (Halliburton Energy Services Group) | Hollingsworth, Malcolm T. (Halliburton Energy Services Group)
Dimensionality reduction in advanced data mining is often a non-linear problem, and the method used to resolve the dimensionality reduction will typically include:
In practice, a data-driven approach that maximizes the sample pair distance match of the original data and the transformed data is usually used. But, the sample mapping optimization and new data transformation are often determined in diverse domains.
This paper provides a hybrid method to determine the appropriate data transformation from an original HD space to an output LD space, usually 2D or 3D space, with minimal information loss. This method is fundamentally a cooperative optimization algorithm combining more than one computational intelligence paradigm. Basically, given the problem data in a HD space, the proposed method first applies evolutionary computation (EC) to determine the LD output values, and then, performs particle swarm optimization (PSO) to refine the results. The data conversion scheme is implemented by a neural network ensemble using the EC-PSO-derived LD outputs as training targets. This method has better capability to tackle problems of local minima and to produce robust conversion of the new data. To make the preserved information essential to the user, multi-objective fitting for advanced data exploration is embedded into this method.
In this paper, both the theoretical background of the developed method and the procedures applied to high-dimensional data visualization, feature extraction, and cluster analyses are discussed. Two case studies with simulated and field data to demonstrate potential applications in reservoir characterization and predictive modeling are also presented. The results strongly justify using a cooperative optimization approach to improve data mapping (especially in handling large data sets), and suggest that the method shows promise as being an effective standard procedure to help automate high-dimensional data processing.
High-dimensional data exploration is crucial for many applications in the oil and gas industry. To effectively facilitate data mining with a large number of variables, some essential information often must be visualized in the data, and more preferably, in a perceptible LD space. In a data-driven approach, essential information typically includes all sample-pair distances for the given data set, and the appropriate dimensionality reduction particularly requires a close match of the distance matrices between the original data and output data. Since the conventional linear mapping methods, principal component analysis (PCA) for example, cannot preserve such distance-based essential information in a satisfactory way, and the optimal position of the data in the output space cannot be obtained analytically in general, determination of the mapping from a HD to a LD space is often treated as a non-linear optimization problem. In practice, given the HD samples, the LD data positions can be optimized initially according to the objective performance criteria selected by the user. Once the LD data positions are determined, many existing methods can be used to model the mapping function based on the available HD and LD examples.