4D Seismic History Matching Research Articles

4D seismic history matching

Reservoir simulation models are used to forecast future reservoir behavior and to optimally manage reservoir production. These models require specification of hundreds of thousands of parameters, some of which may be determined from measurements along well paths, but the distance between wells can be large and the formations in which oil and gas are found are almost always heterogeneous with many geological complexities so many of the reservoir parameters are poorly constrained by well data. Additional constraints on the values of the parameters are provided by general geologic knowledge, and other constraints are provided by historical measurements of production and injection behavior. This type of information is often not sufficient to identify locations of either currently remaining oil, or to provide accurate forecasts where oil will remain at the end of project life. The repeated use of surface seismic surveys offers the promise of providing observations of locations of changes in physical properties between wells, thus reducing uncertainty in predictions of future reservoir behavior. Unfortunately, while methodologies for assimilation of 4D seismic data have demonstrated substantial value in synthetic model studies, the application to real fields has not been as successful. In this paper, we review the literature on 4D seismic history matching (SHM), focusing discussions on the aspects of the problem that make it more difficult than the more traditional production history matching. In particular, we discuss the possible choices for seismic attributes that can be used for comparison between observed or modeled attribute to determine the properties of the reservoir and the difficulty of estimating the magnitude of the noise or bias in the data. Depending on the level of matching, the bias may result from errors in the forward modeling, or errors in the inversion. Much of the practical literature has focused on methodologies for reducing the effect of bias or modeling error either through choice of attribute, or by appropriate weighting of data. Applications to field cases appear to have been at least partially successful, although quantitative assessment of the history matches and the improvements in forecast is difficult.

Accounting for model errors of rock physics models in 4D seismic history matching problems: A perspective of machine learning

Model errors are ubiquitous in practical history matching problems. A common approach in the literature to accounting for model errors is to treat them as random variables following certain presumed distributions. While such a treatment renders algorithmic convenience, its underpinning assumptions are often invalid. In this work, we adopt an alternative approach, and treat model-error characterization as a functional approximation problem, which can be solved using a generic machine learning method. We then integrate the proposed model-error characterization approach into an ensemble-based history matching framework, and show that, with very minor modifications, existing ensemble-based history matching algorithms can be readily deployed to solve the history matching problem in the presence of model errors.To demonstrate the efficacy of the integrated history matching framework, we apply it to account for potential model errors of a rock physics model in 4D seismic history matching applied to the full Norne benchmark case. The numerical results indicate that the proposed model-error characterization approach helps improve the qualities of estimated reservoir models, and leads to more accurate forecasts of production data. This suggests that accounting for model errors from a perspective of machine learning serves as a viable way to deal with model imperfection in practical history matching problems.

4D seismic history matching: Assessing the use of a dictionary learning based sparse representation method

It is possible to improve oil-reservoir simulation models by conditioning them on 4D seismic data. Computational issues may arise related to both storage and CPU time due to the size of the 4D seismic dataset. An approach to reducing the computational requirements is to use a sparse representation method, e.g., Dictionary Learning, to select only the main features of the 4D seismic data. However, the introduction of a sparse representation method incurs an additional computational cost. Alternatively, if one uses ensemble-based methods, it is possible to reduce storage and CPU time by projecting the full seismic dataset on a smaller subspace. This paper evaluates the potential of sparsely representing the seismic data. We compare two experiments, one where we condition on the full dataset projected on a smaller subspace, and one where we use Dictionary Learning to represent the data sparsely. We use Dictionary Learning both on the complete 4D seismic dataset and also on a denoised version of the data. We perform the data assimilation in a slightly different formulation of the Iterative Ensemble Smoother Regularized Levenberg–Marquardt together with correlation-based adaptive localization. We apply these methods to the Brugge benchmark case. Experiment results show that sparse representation methods lead to a final ensemble that is closer to the reference solution, and denoising the seismic data before applying the sparse representation allows us to capture the 4D effect better. Thus, using a sparse representation method in 4D-seismic history matching leads to improved results compared to what we obtain when conditioning the models on the projected 4D seismic dataset.

Technology Focus: History Matching and Forecasting (April 2018)

Technology Focus Today, I am going to talk about abduction. But this is not about kidnapping or a fourth kind of extraterrestrial encounter. I am referring to the meaning of abduction in philosophy literature, where it stands for inference of the best explanation. This topic was brought to my attention by my friend Régis Romeu, former editor of this feature. We use abduction in everyday reasoning when we seek the simplest explanation for an observation. This holds similarities with the current practice of reservoir modeling. In my experience, geoscientists are trained to treat the construction of a model as a process of discovery, a process of finding the simplest, and hopefully the best, explanation for the trends and observations from static and dynamic data. Throughout this process, the geoscientist is “falling in love” with the model. It seems that, instead of the likelihood, the geoscientist’s objective function is based on the “lovelihood.” I am not laying the groundwork for criticizing the work of geoscientists. On the contrary, I have seen colleagues build impressive models that were the foundations of billion-dollar decisions made in highly successful exploration-and-production enterprises. Nevertheless, I believe that it is worth making a counterargument to this process. (Here, I’m taking the liberty of calling the process “model abduction.”) The natural result of model abduction is a base-case model. One unfortunate consequence of a base-case model, however, is the risk of an anchoring effect, in which case we may underestimate uncertainty. Essentially, the anchoring effect refers to our tendency to rely too heavily on the information offered, introducing a bias in the model-construction process. In practice, we might neglect relevant scenarios that were not revealed at the time of the model construction because they did not fit the idea of best explanation. Sometimes, this is referred to as base-case disease. Hence, an alternative approach for model construction is based on generating diversity as opposed to focusing on the best explanation. In this sense, the two approaches seem conflicting. My goal is not to start a controversy, and I am definitely not qualified to state which strategy is best for reservoir modeling. I do see good arguments in both approaches. But, I will venture to say that the best strategy is the one that takes the best of each approach and makes them complementary rather than contradictory. Last year, we had a good number of exciting papers on history matching and forecasting presented at SPE events. The papers summarized in this feature and the ones indicated in the additional-reading list are excellent examples. Interestingly, we had a particularly large number of papers on 4D-seismic history matching. I selected one to be summarized. I hope you enjoy the reading. Recommended additional reading at OnePetro: www.onepetro.org. SPE 185877 Data Analysis Used in Multiple-Realization Work Flows for History Matching—A North Sea Case Study by R. Schulze-Riegert, Schlumberger, et al. SPE 185800 A Simultaneous Bayesian Estimation of Channelized Facies and Reservoir Properties Under Prior Uncertainty by Yu Zhao, The University of Tulsa, et al. SPE 186024 A Variant of Particle-Swarm Optimization for Uncertainty Quantification by V. Babin, Gazpromneft NTC, et al.

Correlation-Based Adaptive Localization With Applications to Ensemble-Based 4D-Seismic History Matching

Summary Ensemble-based history-matching methods have received much attention in reservoir engineering. In real applications, small ensembles are often used in reservoir simulations to reduce the computational costs. A small ensemble size may lead to ensemble collapse, a phenomenon in which the spread of the ensemble of history-matched reservoir models becomes artificially small. Ensemble collapse is not desired for an ensemble-based history-matching method because it not only deteriorates the capacity in uncertainty quantification, but also forces the ensemble-based method to later stop updating reservoir models. In practice, distance-based localization is thus introduced to tackle ensemble collapse. Distance-based localization works well in many problems. However, one prerequisite in using distance-based localization is that the observations have associated physical locations. In certain circumstances with complex observations, this may not be true, and it thus becomes challenging to apply distance-based localization. In this work, we propose a correlation-based adaptive localization scheme that does not rely on the physical locations of the observations. Instead, we use the spatial distributions of the correlations between model variables and the corresponding simulated observations. In the course of history matching, we update model variables by only using the observations that have relatively high correlations with them, while excluding those that have relatively low correlations. This is equivalent to introducing a data-selection procedure to the history-matching algorithm. As a result, the threshold values for data selection play an essential role in the proposed adaptive localization scheme, and we develop both ideal and practical approaches to the choices of the threshold values. We demonstrate the efficacy of the proposed localization scheme using seismic history-matching problems—one 2D and one 3D—in which ensemble collapse is severe in the presence of large amounts of observational data, but distance-based localization may not be applicable because of the lack of physical locations of the seismic data in use. In contrast, correlation-based localization works well to prevent ensemble collapse and also renders good history-matching results. We also note some practical conveniences of the proposed localization scheme, including the applicability to nonlocal observations, the relative simplicity in implementation, the transferability of the same codes among different (either 2D or 3D) case studies, and the adaptivity to different types of observations and petrophysical parameters.

Binary 4D seismic history matching, a metric study

This paper explores 4D seismic history matching and it specifically focuses on the objective function used during the optimisation with seismic data. The objective function is calculated by using binary maps, where one map is obtained from the observed seismic data and the other is from one realisation of the optimisation algorithm from the simulation model. In order to decide which set of parameters is a relevant update for the simulation model, an efficient way is required to measure how similar these two binary images are, during their evaluation within the objective function. Behind this aspect of quantification of the similarities or dissimilarities lies the metric notion, or the art of measuring distances. Four metrics are proposed with this study, the well-known Hamming distance, two widely used metrics, the Hausdorff distance and Mutual Information and a recent metric, called the Current Measure Metric. These metrics will be tested and compared on different case scenarios, designed in accordance to a real field case (gas exsolution) before being used in the second part of the paper. Despite its simplicity, the Hamming distance gives positive results, but the Current Measure Metric appears to be a more efficient choice to cover a wider range of scenarios, these conclusions remain true when tested on synthetic and real dataset in a history matching exercise. Some practical aspects of binary map processes will be examined through the paper, as it is shown that it is more proper to use a derivative free optimisation algorithm and a proper metric should be more inclined to capture global features than local features.

Probabilistic Reservoir-Property Modeling Jointly Constrained by 3D-Seismic Data and Hydraulic-Unit Analysis

SummaryWe present a new method for seismic reservoir characterization and reservoir-property modeling on the basis of an integrated analysis of 3D-seismic data and hydraulic flow units, and apply it to an example of a producing reservoir offshore Western Australia. Our method combines hydraulic-unit analysis with a set of techniques for seismic reservoir characterization including rock-physics analysis, Bayesian inference, prestack seismic inversion, and geostatistical simulation of reservoir properties. Hydraulic units are geologic layers and zones characterized by similar properties of fluid flow in porous permeable media, and are defined at well locations on the basis of logs, core measurements, and production data. However, the number of wells available for hydraulic-unit analysis is often extremely limited. In comparison, the lateral coverage and resolution of 3D-seismic data are excellent, and can thus be used to extend hydraulic-unit analysis away from well locations into the full 3D reservoir volume. We develop a probabilistic relationship between optimal 3D-seismic-data attributes and the hydraulic units that we determine at well locations. Because porosity and permeability distributions are estimated for each hydraulic flow unit as part of the process, we can use the 3D-seismic probabilistic relationships to constrain geostatistical realizations of porosity and permeability in the reservoir, to be consistent with the flow-unit analysis. Reservoir models jointly constrained by both 3D-seismic data and hydraulic flow-unit analysis have the potential to improve the processes of reservoir characterization, fluid-flow performance forecasting, and production data or 4D-seismic history matching.

Time-Lapse Seismic Data Integration Incorporates Pressure and Saturation Effects

This article, written by Special Publications Editor Adam Wilson, contains highlights of paper SPE 166395, ’Streamline-Based Time-Lapse Seismic-Data Integration Incorporating Pressure and Saturation Effects,’ by Shingo Watanabe, SPE, Jichao Han, SPE, Akhil Datta-Gupta, SPE, and Michael J. King, SPE, Texas A&M University, prepared for the 2013 SPE Annual Technical Conference and Exhibition, New Orleans, 30 September-2 October. The paper has not been peer reviewed. This paper presents an efficient history matching approach that simultaneously integrates 4D repeat seismic surveys with well production data. This approach is particularly well-suited for the calibration of the reservoir properties of high-resolution geologic models because the seismic data are areally dense but sparse in time, while the production data are continuous in time but averaged over interwell spacing. The joint history matching is performed using streamline-based sensitivities derived from either finite-difference or streamline-based flow simulation. Introduction Previous studies have made attempts at quantitative 4D seismic history matching. However, the reconciliation of reservoir model heterogeneity with temporal changes in seismic attributes remains a particularly complex task. Several dynamic data integration algorithms have been proposed in the literature, which can be categorized broadly into three data integration levels: (1) reservoir-simulation- based integration between the pressure and saturation estimates inverted from seismic observation data and the direct simulated saturation and pressure responses, (2) petroelastic-based integration between the seismic inverted rock elastic properties derived from a geophysical inversion and the simulated rock elastic responses from the simulated saturation and pressure responses by means of petroelastic models, and (3) seismic-forward-model-based integration between the direct seismic traces attributes and the simulated seismic responses by means of seismic wave propagation modeling. The seismic-forward-model-based seismic data integration approach uses direct seismic trace data and circumvents the uncertainties associated with the seismic-inversion process. The general workflow for generating synthetic seismic data from a reservoir simulation model generally involves (1) static reservoir properties on the simulation grid, (2) simulation of dynamic pressure and fluid saturations at each cell, (3) computation of seismic elastic properties, and (4) simulation of the seismic attribute by applying a seismic wave propagation model over the reservoir interval and the overburden rock. This study examines both reservoir-simulation- based calibration and rock-physics- based calibration. In the first instance, the authors calibrate against time-lapse changes in both pressure and saturation. In the second instance, they calibrate against the time-lapse change in acoustic impedance. In both cases, the inversion will provide local updates to the reservoir model using analytic streamline- based sensitivity calculations.

Sensitivity study of rock-physics parameters for modeling time-lapse seismic response of Norne field

Time-lapse seismic modeling is an important step in joint inversion of time-lapse seismic and production data of a field. Rock-physics analysis is the basis for modeling the time-lapse seismic data. However, joint inversion of both types of data for estimation of reservoir parameters is highly nonlinear and complex with uncertainties at each step of the process. So it is essential, before proceeding with large-scale history matching, to investigate sensitive rock-physics parameters in modeling the time-lapse seismic response of a field. We used the data set of the Norne field to investigate sensitive parameters in time-lapse seismic modeling. We first investigated sensitive parameters in the Gassmann’s equation. The investigated parameters include mineral properties, water salinity, pore pressure, and gas-oil ratio. Next, we investigated parameter sensitivity for time-lapse seismic modeling of the Norne field. The investigated rock-physics parameters are clay content, cement fraction, average number of contact grains per sand, pore pressure, and fluid mixing. We observed that the average number of contact grains per sand had the most impact on time-lapse seismic modeling of the Norne field. The clay content was the most sensitive parameter in fluid substitution for calculating seismic velocities of the Norne field. Salinity and pore pressure had minimal impact on fluid substitution for this case. This sensitivity analysis helps to select important parameters for time-lapse (4D) seismic history matching, which is an important aspect of joint inversion of production and time-lapse seismic modeling of a field.

4D Seismic history matching using information from the flooded zone

We introduce a simplified way of using 4D seismic amplitude differences, without explicit full-physics modelling, in a history matching workflow that includes conditioning to both well production data and flooded-zone patterns derived from time-lapse seismic surveys. This approach reduces the complexity of the integration workflow, and speeds up the integration cycle. Examples from both synthetic and field cases demonstrate the success of the workflow.

Stochastic Data Integration: History Matching Production and Time-Lapse-Seismic Data

This article, written by Senior Technology Editor Dennis Denney, contains highlights of paper SPE 146418, ’A Comparison of Stochastic Data-Integration Algorithms for the Joint History Matching of Production and Time-Lapse-Seismic Data,’ by Long Jin, SPE, Faruk O. Alpak, SPE, and Paul van den Hoek, SPE, Shell International E&P, and Carlos Pirmez, Tope Fehintola, SPE, Fidelis Tendo, and Elozino Olaniyan, SPE, Shell Nigeria E&P, prepared for the 2011 SPE Annual Technical Conference and Exhibition, Denver, 30 October-2 November. The paper has not been peer reviewed. Quantitative integration of spatial and temporal information provided by time-lapse (4D) -seismic surveys into dynamic reservoir models requires efficient and effective data-integration algorithms. The particle-swarm-optimization (PSO) method emerged as more effective compared with the neighborhood algorithm (NA) and very-fast-simulated-annealing (VFSA) methods in the Imperial College Fault Model (ICFM) problem. The PSO method was also effective in the field application. The VFSA method required comparatively more iterations to converge because of its sequential nature, but it has advantages when moderate computing resources are available. Introduction The use of 4D-seismic data is increasingly important for reservoir monitoring and management. However, quantitative integration of 4D-seismic data with historical production data into reservoir-simulation models is a challenging task. A key issue of joint history matching, involving production and time-lapse-seismic data, is proper processing of 4D-seismic data such that physical information (fluid saturations) is captured by geophysical measurements properly. Various types of geophysical measurements have been assimilated in the literature, including inverted impedance, inverted saturation, and fluid fronts. Another important issue centers on appropriate selection of model parameters to be adjusted to obtain a history match. In this work, three stochastic data-integration methods were evaluated: NA, VFSA, and PSO. They were compared on the joint history-matching problem involving production and time-lapse-seismic data. Stochastic Data-Integration Methods Data integration in the presence of time-lapse-seismic data consists of updating reservoir dynamic models by jointly assimilating both 4D-seismic and production data. The goal is to use the complementary nature of 4D-seismic data (spatial and temporal information) and production data (temporal information) to produce more-predictive reservoir models, and in turn, make optimal reservoir-management decisions. In this work, the focus was on nonintrusive stochastic data-integration methods. No special information (e.g., adjoint sensitivity coefficients or gradients of the objective function) was needed from the flow and seismic forward models. A generic iterative workflow for joint 4D-seismic and production history matching conducted by a stochastic sampling engine (optimizer) is shown in Fig. 1. The main idea was to perturb the uncertain (model) parameters identified by comparing the synthetic data (production and 4D-seismic) with the measurements simultaneously. The basic idea is also shown in Fig. 2.

Data-Partitioning Technique for Gradient-Based-Optimization Methods in History Matching

This article, written by Senior Technology Editor Dennis Denney, contains highlights of paper SPE 130473, ’Development of a Data-Partition Technique for Gradient- Based-Optimization Methods in History Matching,’ by D.Y. Ding, SPE, IFP, prepared for the 2010 SPE EUROPEC/EAGE Annual Conference and Exhibition, Barcelona, Spain, 14-17 June. The paper has not been peer reviewed. Assisted history matching is used widely to constrain reservoir models. However, history matching is a complex inverse problem, and it is always a challenge to history match large fields that have a large number of parameters. A new technique is presented for gradient-based-optimization methods to improve history matching of large fields. This technique uses data partitioning for the gradient calculations. This perturbation design enables calculating all derivatives of the objective function with only a few perturbations. Introduction Assisted history matching constrains reservoir and/or geological models by integrating well-production data and/or 4D-seismic data. The degree of difficulty and the computational effort (in terms of the number of reservoir simulations) increase with the increasing number of matching parameters. Generally, a reasonable history match is acquired at the field level first, then at the regional level, followed by more-rigorous individual-well history matching. The number of parameters usually is increased when the history matching moves to the regional or well levels. Therefore, it is important to study optimization methods for a large number of parameters, especially for large fields. Among optimization methods for history matching, gradient-based approaches are prevalent. However, gradients of the objective function often are calculated by numerical methods, which need the evaluation of the objective function through reservoir simulations, which, in turn, need considerable CPU time. To optimize M parameters, at least M perturbations (M+1 simulations) are required to calculate all the gradients to obtain an optimized solution. When M is large, serious problems in CPU time can be encountered. The adjoint method is an efficient technique to reduce the number of simulations for gradient calculation. However, an accurate formulation and solution of the adjoint method requires specific knowledge of the discrete equations in the reservoir simulator. In addition, this method is not always computationally feasible for constructing the Hessian matrix for the Gauss-Newton approach in the optimization. An alternative approach for reducing the number of simulations in history matching is to split the objective function and/or parameters by subregions. One technique uses principal-component analysis for the Hessian matrix to delineate multiple reservoir regions for history matching. Another method uses independent objective functions with a direct-search algorithm. A polynomial-response surface defines separable subregions in parameter space and applies the neighborhood algorithm for the optimization in 4D-seismic history matching. However, full independence of subregions within a reservoir generally is not possible. Interaction between different regions should be considered.

Time-lapse seismic inversion for pressure and saturation in Foinaven field, west of Shetland

Christophe Ribeiro and Colin MacBeth present a new petro-elastic-based approach to independently estimate reservoir pressure and saturation from 4D seismic data obtained at BP’s Foinaven field, west of Shetland. In the last decade, the use of seismic monitoring (4D) has greatly increased, and oil companies have shown their commitment to this technology (Marsh et al., 2003; De Waal and Calvert, 2003), as confirmed by its more systematic application. Time-lapse seismic is now an integrated part of reservoir management strategy, and has proved its ability to improve reservoir understanding. In the past, the applications of seismic monitoring were predominantly focused on the tracking of fluid contacts and movements (i.e. gas-cap expansion, water sweep); discrimination between lithology and fluid effects; and identification of pressure compartment and reservoir connectivity. The growth of seismic reservoir monitoring has brought new technical challenges to the industry, in order to improve and enlarge the existing portfolio of this technology. In fact, the advance in acquisition (i.e. steerable streamers, full-wave field recording devices, permanent sensors), survey design (i.e. 4D dedicated acquisition), seismic processing (i.e. specific 4D workflow) and integration between contractors and oil companies (i.e. in-house processing and R&D teams) have improved the repeatability and quality of the data, and paved the way for new applications. Research has now moved towards the more quantitative aspects of 4D, attempting to estimate hydrocarbon-production-related changes directly from seismic data. However, in most cases, seismic amplitude changes are not only due to variation in fluid saturation, pore pressure, or even compaction, but to a combination of these effects. The goal of distinguishing between pore pressure and fluid saturation effects is a challenge in quantitative time-lapse seismic interpretation, and has been the subject of many papers. The separation of pressure and saturation effects has mainly been tackled by means of rock-physics-based techniques (Brevik, 1999; Tura and Lumley, 1999; Landrø, 2001), and recently engineering approaches using production, pressure and PVT data, have been developed (He et al., 2004; MacBeth et al., 2004). The independent estimation of pore pressure and fluid saturation, which are also two products from the reservoir flow simulation, allows a direct comparison between attributes derived from the seismic and engineering domains. These production attributes could be used in a 4D seismic history matching workflow in order to directly constrain the predictions from the fluid flow simulator with the seismically derived dynamic properties (Gosselin and Menezes, 2006; Stephen and MacBeth, 2006). Modifications of the reservoir model could also be made to improve the agreement between the flow simulation results and the seismically derived production estimates, and ultimately to increase reservoir performance and oil recovery.