Search

With the enormous increase in the demand for crude oil, and decrease in the resources of conventional oil reservoirs, there is a great need to understand heavy or foamy oil-gas drive mechanism to maximize the oil and gas production. To analyze the real movement of non-viscous heavy oil flow, the characteristic features of the oil-gas mixture has to be estimated to forecast the future potential supply from a heavy oil reservoir. An important question in heavy oil flow under solution gas drive is whether the behaviour of depletion tests can be simulated to model the heavy oil flow behaviour. The main objective of this research is to develop a reliable numerical model for modelling heavy oil flow calibrated with controlled solution gas drive experiments, and that makes a novelty in this manuscript. In this paper, CMG-STARS model which is capable of simulating solution gas drive tests that matched the research experiments. This heavy oil recovery model can determine the relative permeability curves for oil and gas in the dual-phase system using Corey’s relations. At a depletion rate of 0.0418 psi/min, the maximum cumulative oil and gas production was observed to be 13,000 cm3 and 8500 cm3, respectively. The results from the bottom hole pressure and the block pressure simulation runs indicate that the fluid properties such as surface tension plays a significant role in the gas bubble formation. These results are promising, and helps to understand the complex behaviour of heavy oil reservoirs and thus can improve heavy oil recovery.

Sand production is a problem that affects hydrocarbon production from unconsolidated sandstone reservoirs. Several factors, such as the strength of the reservoir, its lithification and cementation and reduction in pore pressure, may cause sand to be separated from the rock and transported by hydrocarbons to the well. Producing sand commonly causes erosion and corrosion of downhole and surface equipment, leading to production interruptions and sometimes forces operators to shut-in wells. Several different methods of sand control are available to reduce the impact of sand production. The reviewed papers suggest that the most suitable methods for unconsolidated sandstone reservoirs are stand-alone screens and gravel packs. Because of the cost and complexity of gravel packs, stand-alone screens are usually the first choice. These screens have different geometries, and selection of the most suitable screen depends on the particle size distribution of the grains in the formation and other reservoir and production parameters. A screen retention test, run in a laboratory with screen samples and typical sands, is often used to ensure that the screen is suitable for the reservoir. This paper reviews the main causes of sand production, the properties of unconsolidated sandstones that predispose reservoirs to sand production problems and the selection criteria for the most suitable mitigation method. The process of selecting a screen using experimental screen retention tests is reviewed, and the limitations of these tests are also discussed. Some numerical simulations of experimental tests are also reviewed, since this represents a very cost-effective alternative to laboratory experiments.

Oil and gas exploration gradually changes to the deep and complex areas. The quality of seismic data restricts the effective application of conventional time-frequency analysis technology, especially in the case of low signal-to-noise ratio. To address this problem, we propose a curvelet-based time-frequency analysis method, which is suitable for seismic data, and takes into account the lateral variation of seismic data. We first construct a kind of curvelet adapted to seismic data. By adjusting the rotation mode of the curvelet in the form of time skewing, the scale parameter can be directly related to the frequency of the seismic data. Therefore, the curvelet coefficients at different scales can reflect the time-frequency information of the seismic data. Then, the curvelet coefficients, which represent the dominant azimuthal pattern, are converted to the time-frequency domain. Since the curvelet transform is a kind of sparse representation for the signal, the screening process of the dominant coefficient masks most of the random noise, which enables the method to adapt for the low signal-to-noise ratio data. Results of synthetic and field data experiments using the proposed method demonstrate that it is a good approach to identify weak signals from strong noise in the time-frequency domain.

Fluvial sediments with multi-scale channels are difficult to model using classical two-point statistical methods, e.g. sequential indicator simulation (SIS) or object-based modelling (ObjM). Multiple-point statistics (MPS) has been used to generate facies, fracture and porosity distributions based on pattern statistics derived from training datasets. However, the ability of these three methods to reproduce channel geometry and continuity is not clear, especially when using differently spaced conditional data. This paper presents a case study to compare the application of these three methods in reproducing channels from a section of Amazon River based on two differently spaced conditional data sets. Results show that: the reproduction accuracy is similar between MPS and SIS; MPS provides the most connected channel facies (or most channel continuity) as compared to SIS and ObjM; and using a hand-drawn facies based on the sampling points yield a similar accuracy to that achieved by using the reality facies distribution as the training image. Finally, we conclude that the application of MPS does not significantly increase the reproduction accuracy when compared to SIS channel models; however, MPS can generate realistic models with respect to channel geometry and continuity.

Understanding the controls on coal seam distribution and geometry is fundamental for planning coal seam gas production. In the Jurassic Surat Basin of South East Queensland, Australia, the spatial continuity of coal seams in the Walloon Coal Measures is highly variable and often difficult to map and predict, even with closely spaced (

A surrogate model approximates a computationally expensive solver. Polynomial Chaos is a method used to construct surrogate models by summing combinations of carefully chosen polynomials. The polynomials are chosen to respect the probability distributions of the uncertain input variables (parameters); this allows for both uncertainty quantification and global sensitivity analysis. In this paper we apply these techniques to a commercial solver for the estimation of peak gas rate and cumulative gas extraction from a coal seam gas well. The polynomial expansion is shown to honour the underlying geophysics with low error when compared to a much more complex and computationally slower commercial solver. We make use of advanced numerical integration techniques to achieve this accuracy using relatively small amounts of training data.

We present a new computational method of automatic normal moveout (NMO) correction that not only accurately flattens and corrects the far offset data, but simultaneously provides NMO velocity ( $$v_\mathrm{nmo}$$ ) for each individual seismic trace. The method is based on a predefined number of NMO velocity sweeps using linear vertical interpolation of different NMO velocities at each seismic trace. At each sweep, we measure the semblance between the zero offset trace (pilot trace) and the next seismic trace using a trace-by-trace rather than sample-by-sample based semblance measure; then after all the sweeps are done, the one with the maximum semblance value is chosen, which is assumed to be the most suitable NMO velocity trace that accurately flattens seismic reflection events. Other traces follow the same process, and a final velocity field is then extracted. Isotropic, anisotropic and lateral heterogenous synthetic geological models were built to test the method. A range of synthetic background noise, ranging from 10 to 30 %, was applied to the models. In addition, the method was tested on Hess’s VTI (vertical transverse isotropy) model. Furthermore, we tested our method on a real pre-stack seismic CDP gathered from a gas field in Alaska. The results from the presented examples show an excellent NMO correction and extracted a reasonably accurate NMO velocity field.

The demand for natural gas in Queensland, Australia has historically been supplied from conventional reservoirs. However, depletion in conventional sources has led producers to turn to extensive supplies in Queensland's coal resources. These coal seam gas (CSG) developments not only represent new supplies for the domestic market in eastern Australia, they are also the first time that CSG (aka coal bed methane or CBM) has been liquefied to serve the expanding world LNG market. In order to make this development occur, considerable infrastructure had to be installed, with field developments still on-going. This AUD$60 billion investment precipitated a major overhaul of state regulations to provide not only a safe and clean operating environment, but also to allay the concerns of certain stakeholders. The gas is primarily produced from thin high permeability coals in the Jurassic-age Walloon Coal Measures in the Surat Basin and from several relatively thick Permian-age coal seams in the Bowen Basin, of which the Baralaba Coal Measures and the Bandanna formation are the most important. There are numerous technical challenges with this production, such as fines production from the inter-burden clays, which can form a thick paste that is difficult to pump. Salt extraction by reverse osmosis, from associated water produced to depressurise the coal seams and enable the flow of gas, allows for the beneficial use of the water. Technical challenges also include mathematical modelling of the counter-current two-phase flow (gas and water) in the well annuli because conventional models in simulators only handle co-current two-phase flow in the well-bores. Also, the subject of on-going investigations is decommissioning of the large number of shallow wells over the next few decades in a safe and cost effective manner, with compressed bentonite being a promising option for well plugging. As with any major commercial development, in addition to the technical challenges there have been social challenges as well. These include interaction and coexistence of extensive surface operations with an established agricultural sector, interactions between gas production and ground water aquifers in water-stressed areas, and the cumulative social and economic impacts of 3 large projects on a rural area. Ultimately, the State of Queensland expects to produce more than 1800 BCF/annum, of which about 1400 BCF/annum will be exported as LNG. Depending on the demand and well productivity, up to 1000 CSG wells may be drilled per year for the next thirty years. A review of CSG resources, development, and challenges is presented in this paper to provide context for a stream of research findings that are emerging on the Queensland CSG experience.

Spatial coalbed methane (CBM) resource estimation is based on spatial distributions of coal, coal adsorbed gas content and coal density. However, the spatial distribution of gas content can be generated via two different geological modeling processes: (1) The gas content distribution is generated by geological modeling based on the interpreted gas content at boreholes; (2) distributions of gas content related logs or coal properties are generated firstly, then the gas content distribution is calculated based on the spatial distributions of logs or coal properties by the relationship between the gas content and logs or coal properties. This paper presents a study to compare the impact of these two processes on CBM resource estimation for coal seam no. 3 (CS-3) in southeast Qinshui Basin, China. Well logs from 22 wells, laboratory data from five wells and well tops from 131 wells for CS-3 are used in log interpretation and geological modeling. The simple kriging (SK) is used to build the structural model and the coal distribution. Weighted and unweighted omni-directional variograms for structural residual and coal thickness are calculated using an in-house program. Logs of gamma-ray (GR) and density (DEN or RHOB) are distributed in 3D by using sequential Gaussian simulation (SGS) with SK algorithm. Artificial neural network (ANN) is used to build the relationship of the measured raw gas content (RGC; gas content in raw coal basis) with the logs of GR, DEN and measured depth (MD). Then the RGC is distributed in 3D by the two geological modeling processes. CBM resources are calculated in 3D based on the cells' volume, coal density and RGC. Results show that RGC increases with an increase in burial depth. Total CBM resources for the study area calculated by these two processes are similar for CS-3 but the distribution probability of high gas content is highly different which is important for locating wells.

The development of effective algorithms for the simulation of facies in petroleum systems continues to be limited to techniques relying on linear spatial dependence; Truncated Gaussian Simulation, Plurigaussian Simulation and Sequential Indicator Simulation, in addition to pattern-based techniques based on Multiple Point Statistics. Recent research in the development of new geostatistical techniques for continuous variables has found that the requirement of linear spatial dependence is very rarely satisfied. Datasets from micrometre to kilometre scale have been analysed spatially and, in all cases, significant non-linearity in the spatial dependence structure have been observed. This paper replaces the Gaussian ‘engine’ in Truncated Gaussian Simulation and Plurigaussian Simulation, which requires linear spatial dependence, with a copula-based algorithm that is able to model non-linear dependence structures. The advantage is a facies model that does not contain the symmetric spatial artefacts that characterise Gaussian-based models. In flow simulation terms, this means that the connectivity (and directional connectivity) of different facies within the reservoir architecture to be modelled with greater fidelity, honouring the expected continuities of each of the facies elements. Furthermore, extending recent research by Horning and Bardossy on directional and order asymmetry for continuous variables to consider their application to facies models allows the development of facies-models that are able to represent the directional asymmetry found in sedimentary depositional systems. This allows a better representation of the distal and proximal elements in depositional systems as well as the opportunity to model complex asymmetries associated with meandering channel and estuaries.

Coal seam gas (CSG) is naturally occurring methane found in coal seams. Fractures act as the major channel for the gas to flow and play a significant role in evaluating the practicability of the CSG exploitation. Prediction of the subsurface coal seams and the distribution of the fractures within the coal seams is a challenge. This study attempted to predict coal distribution by applying spectral decomposition and post-stack seismic inversion on a data set of the Surat Basin, Queensland, Australia. Semblance and curvature attributes were computed to determine fracture distribution within the coal seams. According to the rock physics analysis, based on well data, coal beds had a low p-impedance as compared to surrounding lithologies. A model-based P-impedance inversion was executed to create a P-impedance volume. Extracted horizon slices successfully highlighted coal seams at different levels. Blind test for the P-impedance prediction at a well location revealed a reasonable match for coal prediction. High frequency (from 40~60 Hz) spectra delineated thin coal beds. Semblance and curvature predicted zones of high fracture density within various stratigraphic levels. The high density of fractures is along faults. This study successfully demonstrated the prediction of coal seams and fracture distribution by using seismic inversion spectral decomposition and seismic attributes. Hence, these techniques can be used for the field development of coal seam gas.

A number of different techniques are currently used in the estimation of petrophysical properties away from hard data locations using a softer secondary data such as a seismic attribute. There are two problems that affect these techniques; firstly the relationship between the primary and secondary data is often very weak and secondly the measurement scale of the primary and secondary data is significantly different. Seismic data is usually well represented laterally but has a vertical resolution almost an order of magnitude larger than the well log data that is used as the primary data source.Currently the relationship between primary and secondary variable is quantified by a correlation coefficient and possibly a cross-variogram. Neither of these two measures capture the scale variability of the relationship or any location information between the two property fields. This paper investigates the use of the continuous wavelet transform as a way of examining the scale relationship between these two data sources with the aim of improving the understanding of the behaviour of primary data at different scales. The technique appears to indicate that there is also a potential improvement in the correlation between primary and secondary data once an appropriate comparison scale has been identified.This more robust method for determining the scale dependent relationship between primary and secondary data suggests a new estimation workflow that reduces the uncertainty associated with the estimation of primary data using secondary data sources.

Since computer modelling of petroleum reservoirs was developed in the late 1980's, the technique has become almost ubiquitous. But, most of the experienced geoscientists that really understand the capabilities of these models are sceptical about the predictive capability of reservoir modelling and acknowledge that even their best models are either highly subjective or are quantified guesses. In recent years a trend has been emerging in which reservoir models are increasingly demanded by regulatory authorities prior to development approval which raises the question: "What is a good reservoir model?" This paper considers the concept of a ‘good model’ since, without clear guidelines on good models it is difficult to determine who is qualified to build a model or who is qualified to audit a model. The process of verification and validation of reservoir models is then considered and petroleum industry practices are compared with computer modelling in other industries. There is increasing public concern over the development of unconventional resources, and industry concern over possible litigation, so this paper suggests a set of acceptance criteria for sub-surface models with the aim of identifying procedures that are auditable, defendable and robust for adoption by regulators. During the development of these criteria, a number of key areas in modelling workflows were identified as being of critical concern to the construction of a ‘good model’; model scope, prior elicitation, scale issues, and multiple scenarios. These four areas are discussed in detail and their impact on model quality and predictive capability are assessed. It is hoped that the topics highlighted in this paper encourage enlightened discussion, and ultimately consensus, on the concept of good quality sub-surface modelling of unconventional resources between the regulators and the industry.

Laterally discontinuous coal measures are common in alluvial settings due to interaction with fluvial systems. Under these conditions it is difficult to accurately represent coalbeds and interburden sandstone bodies in static and dynamic models at a regional scale. These challenges are compounded in the Walloon Coal Measures by non-uniform drill spacing, which varies from clustered to sparse and insufficient outcrop exposures available to constrain the correlations. To address these issues, this study investigates a nested approach to facies modelling of the Upper Juandah Member of the Jurassic Walloon Coal Measures in the Surat Basin, Queensland, which contains some 3,600 wells, of which half were analysed for lithofacies distributions. This approach contrasts the application of truncated Gaussian simulation, object modelling and multiple-point geostatistical simulation. First, a regional scale structural model was developed based on the correlation of sub units within the basin and the lithofacies were then interpreted from normalised wireline logs. Then geometries of individual facies were defined from two local scale models (~6 × 6 km2) where dense drilling, 3D seismic and paleocurrent analysis data were available to constrain the models. Three training images, generated by object modelling, an analogue of one part of the Ob River, and an interactive method were subsequently used to model primary channels, channels and crevasse splays, respectively. Truncated Gaussian simulation was used in modelling the distribution of marginal and coal swamp. The final model is a combination of the model with primary channels and channels, and the model with marginal and coal swamp. This approach is the first trial in modelling swamp and channel distributions at a regional scale by integrating data from local models, depositional analogues and paleo-flow interpretation in the Surat Basin.

As time passes our understanding of a reservoir changes, more information becomes available. Our original ideas about the geology, the fault compartments and the ultimate recoverable volume of oil or gas are modified by the acquisition of additional information about the reservoir. Our expectation is that as we collect more data, our uncertainty about the reservoir is reduced. As we know more about the reservoir our expectation is that our predictions for the recoverable hydrocarbons will become more accurate. This paper considers the dilemma of deciding how to incorporate new information into the model. Most of the time, it appears that there is no consistency or clarity in the strategy. More worryingly, it appears that not all data is considered equally.

SUMMARY Fluvial systems host coal seam gas reservoirs in various fields of Queensland. However, the lateral heterogeneity of reservoirs properties within these reservoirs can be significant and determining the distribution of these reservoirs is a challenge. This study attempts to predict coal distribution by applying rock physics and post-stack seismic inversion on data set of Scotia field of the Surat Basin. According to rock physics analysis, coal beds have significantly low density. Consequently, this gives low Pimpedance as compared to surrounding lithology. Therefore, inverted P-impedance and density volumes can be used to predict coal distribution and connectivity of different coal seam reservoirs. Theoretically, density volumes may provide accurate prediction, but this requires execution of comprehensive pre-stack inversion workflow. We only used a model based P-impedance inversion technique to create P-impedance volumes in order to better image the reservoir and connectivity. Extracted horizon slices by using cutoff based on rock physics analysis, successfully highlights architecture of coal beds. Computed average P-impedance within zone of interest can provide information regarding promising zones for exploration of coal seam gas. Blind test for P-impedance prediction at well locations reveals reasonable match for coal prediction using this method.

This paper demonstrates a case study of a hybrid methodology based on the combination of radial basis function neural network and sequential Gaussian simulation. The methodology is demonstrated with an application to modelling the porosity distribution in an oil reservoir of the Lower Tertiary in the north of Dongying depression, Shengli Oilfield, East China. The methodology first uses radial basis function neural networks to estimate the porosity trends (fluvial directions) from high-dimensional data system with both well and seismic data. Gaussian simulation helps to do the local uncertainty analysis for the reservoir model. The final results from the hybrid methodology assure our confidence on the reservoir model both horizontally and vertically. They are realistic and honour the geological rules of the oilfield. The technique is fast and straightforward, and provides an effective computational framework for conditional simulation.

Upscaling techniques developed in the late 1980s continue to form the basis of including detailed geological data in flow simulation models. While these techniques are theoretically robust, in practice the process of upscaling has variable success and the concept of replacing a complex and detailed geological model with an array of effective properties is considered to be a source of error in the construction of simulation models. The default ratio between the geological grid and the simulation grid in popular geological modelling programs is eight fine cells to one coarse cell. Examination of the number of samples required to estimate the mean value of a population suggest that this ratio should be much higher, and that more accurate upscaling results would be obtained if there were 1,000 fine cells in each coarse cell. This suggests that a simulation grid of 500,000 cells should have a corresponding geological grid with 500 million cells to improve the accuracy of the computation of effective properties. This is beyond the capabilities of the present generation of commercial geological modelling software used in the petroleum industry. This paper discusses the creation of gigacell grids, and the ease of construction and difficulties in population using existing geostatistical algorithms.

As 3D reservoir modeling is effectively the centre stage in many multi-disciplinary reservoir management efforts, the need for effective knowledge management is paramount to ensure: Inclusion of pre-existing knowledge is in the model, including absent (or controversial knowledge), reflected in uncertainties. A definition of a modeling process comprehensively covers the reservoirs issues identified while leading to fit for purpose results in a given imposed time span. Storage of knowledge generated and a modeling audit trail during the modeling process to ensure adequate and efficient model updates over time. Creation of a link between the integrated modeling process and the discipline processes that generate the basic data that underpin the model Current 3D modelling methods and tools allow for creation of models and storing model specific workflows (i.e. software related parameters and processing chains). In addition, various efforts are undertaken to store modelling best practice, which typically entails general know-how about creating models. During a specific 3D modelling project, however, a considerable body of knowledge is generated about what and what not makes the model work for a given reservoir. Such knowledge, which are essential for systematic uncertainty assessment during an existing modelling effort or subsequent modelling efforts, needs to be managed in order to retain its relevance. This extended abstract focuses on application of knowledge management in the EP industry, particularly on how process or workflow-based knowledge management approaches add value in the context of 3D modeling projects for multidisciplinary reservoir management.

In recent years the construction of sub-surface models to understand complex systems has become almost ubiquitous. At the same time, societies have become more litigious and governments in the Asia-Pacific region have become more demanding in their technical requirements for exploration and development approvals. But most experienced geoscientists that really understand the capabilities of these models are much more sceptical about the predictive capability of reservoir modelling and acknowledge that even their best models are either highly subjective or quantified guesses. If reservoir models provide technical justification for investment decisions or, especially in the case of CO2 sequestration, decisions that may face future litigation, then it is important to be able to quantify their capabilities and to place reasonable limits on the expectations of these models. The simple question, ‘What is a good reservoir model?’ raises many different answers from a group of geoscientists, and yet if we are not clear on what a good model is, how can we conceptualise better models as distinct from worse models? Without clear guidelines for good models it is difficult to determine who is qualified to build a model or who is qualified to audit a model. In this situation, if authorities continue to increase their demands for models to support reservoir management decisions, then many months of effort will continue to be wasted in the construction of models with no numerical validity or predictive capability. One of the aims of this paper is to open discussion on these issues and to suggest areas where clear consensus needs to be established for the benefit of the industry, the Government and our society. Uncertainty in structural interpretation and uncertainty in deposition have a significant impact on good reservoir management for individual companies and on the wise management of national resources. The application of good modelling practices and appropriate validation by regulatory authorities is a sensible aim for our industry and one that has been overlooked for too long.

A new method for simulating conditional spatial fields is presented which improves on linear geotatistics currently used in commercial petroleum software. It preserves the continuity of extreme high and low value regions and provides an assessment of uncertainty based on the dependence of the magnitude of adjacent values. Moreover, this method includes flexibility to handle a variety of conditioning constraints, including non-linear constraints, integral equalities and inequalities.