Enhanced characterization of reservoir hydrocarbon components using electromagnetic data attributes

Abstract

Advances in electromagnetic imaging techniques have led to the growing utilization of this technology for reservoir monitoring and exploration. These exploit the strong conductivity contrast between the hydrocarbon and water phases and have been used for mapping water front propagation in hydrocarbon reservoirs and enhancing the characterization of the reservoir formation. The conventional approach for the integration of electromagnetic data is to invert the data for saturation properties and then subsequently use the inverted properties as constraints in the history matching process. The non-uniqueness and measurement errors may however make this electromagnetic inversion problem strongly ill-posed, leading to potentially inaccurate saturation profiles. Another limitation of this approach is the uncertainty of Archie's parameters in relating rock conductivity to water saturation, which may vary in the reservoir and are generally poorly known. We present an Ensemble Kalman Filter framework for efficiently integrating electromagnetic data into the history matching process and for simultaneously estimating the Archie's parameters and the variance of the observation error of the electromagnetic data. We apply the proposed framework to a compositional reservoir model. We aim at assessing the relevance of EM data for estimating the different hydrocarbon components of the reservoir. The experimental results demonstrate that the individual hydrocarbon components are generally well matched, with nitrogen exhibiting the strongest improvement. The estimated observation error standard deviations are also within expected levels (between 5 and 10 %), significantly contributing to the robustness of the proposed EM history matching framework. Archie's parameter estimates approximate well the reference profile and assist in the accurate description of the electrical conductivity properties of the reservoir formation, hence leading to estimation accuracy improvements of around 15 %.

SAMA: Spatially-Aware Model-Agnostic Machine Learning Framework for Geophysical Data

“…Methods including Artificial Neural Networks (ANNs) [9], [16]- [20], Support Vector Machine (SVM) [18], [19], and Decision Trees (DT) [19] have been used to predict the saturation from well log data. As predictions based on welllogs can extend only a few meters surrounding the borehole, seismic data [21]- [23], and crosswell electromagnetic surveys [10], [11], [23] are used for the estimates to extend to a few kilometers surrounding the borehole. These approaches allow the engineers to obtain information about the interwell region and provide estimates for the water saturation.…”

“…These methods require knowledge of extra parameters such as permeability or the presence of a physical or numerical model. Other studies use history matching techniques that require several snapshots in time to make the predictions [10], [11]. To the best of our knowledge, there are no studies that rely only on machine learning techniques for water saturation distribution mapping from a single snapshot of Crosswell EM surveys and porosity distribution.…”

Efficient Assimilation of Crosswell Electromagnetic Data Using an Ensemble-Based History-Matching Framework

“…As outlined in Hamada et al (2013), the estimates of Archie's parameters can be heterogeneous and quite uncertain in some applications, especially in the case of carbonate reservoirs. Although these uncertain parameters can be readily incorporated into the history-matching framework (Katterbauer et al, 2016), we neglect this uncertainty in this study and assume that Archie's parameters are homogeneous and already determined from core or log analysis. In Eq.…”

“…There has been a growing interest in enhancing reservoir characterization through history matching using EM observations. Katterbauer et al (2016) presented a history matching study using EM data attribute (conductivity, assumed to be known) to estimate the components of a compositional reservoir model with the EnKF, in which the uncertainty in Archie's parameters and the variance of observation error was also considered. In addition to EM conductivity attribute, Katterbauer et al (2015) also incorporated seismic, gravimetry and InSAR data for history matching using the EnKF to study the synergy effect that could result from combing them.…”

Application of an Adaptive Multi-Population Parallel Genetic Algorithm with Constraints in Electromagnetic Tomography with Incomplete Projections

“…Electromagnetic tomography (EMT) has particular advantages over other exploration methods, such as drilling methods, 3-D seismic exploration, and channel wave tomography, as it can realize high imaging resolution, has convenient construction, and is non-hazardous [2][3][4][5]. EMT has been widely used in the exploration of coal, oil, natural gas, and other mineral resources [6][7][8]. However, in the process of exploration, the ray coverage angle in the observation system is limited by field detection conditions, and the electromagnetic wave projection data obtained are always incomplete.…”