Elsevier

Computers & Geosciences

Volume 128, July 2019, Pages 30-40
Computers & Geosciences

Fast assimilation of frequently acquired 4D seismic data for reservoir history matching

https://doi.org/10.1016/j.cageo.2019.04.001Get rights and content

Highlights

  • We developed and applied a framework for assimilating frequently acquired 4D seismic data to update reservoir models.

  • The proposed method introduces a new attribute for seismic history matching.

  • We deploy Morris sensitivity analysis and ES-MDA to analyze and calibrate uncertainty parameters in reservoir models.

  • The proposed framework shows to boost history matching efficiency on frequently acuquired 4D seismic data.

Abstract

The study in this paper proposed a new framework for history matching of frequently acquired 4D seismic data. To achieve this goal, the large volumes of seismic data from the many repeated 4D monitors are firstly condensed into a single attribute by directly correlating them to the reservoir production and injection performances. This ‘well2seis’ cross-correlation is achieved by defining a linear relationship between pressure and saturation-related 4D seismic responses and the cumulative changes of reservoir fluid volumes derived from wells. It is shown that such a cross-disciplinary attribute not only reduces the amount of 4D seismic data for history matching, but also enhances the seismic data reliability since the 4D seismic observations are conditioned by low-uncertainty production data from reservoir engineering domain. In the second step, Morris sensitivity analysis is adapted to fast diagnose the uncertainty reservoir model parameters that are sensitive to the well2seis attributes. To quantitatively assimilate the well2seis observations to calibrate these uncertainty parameters, we then proposed a well2seis objective function that quantifies the mismatch between the observed and model simulated well2seis attributes. Ensemble Smoother with Multiple Data Assimilation (ES-MDA) is performed at the end to iteratively assimilate the well2seis observations by minimizing the well2seis misfit. Application of the proposed workflow to a North Sea field case shows that, when history matching to the observed well2seis attribute that honours the information from seismic and reservoir engineering domains, it can significantly reduce the uncertainty of key reservoir parameters, hence improving the matching quality to both 4D seismic and production observations and enhancing the prediction reliability of the reservoir models. Compared to traditional history matching approaches that attempt to match individual seismic time-lapse attributes and production observations, this approach is observed to significantly boost the history matching efficiency by reducing the number of time-consuming iterations.

Introduction

4D seismic monitoring is of significant importance for hydrocarbon reservoir surveillance and CO2 sequestration assessment. For instance, 4D seismic data has shown the strongest impact on deep-water developments in West Africa and the Gulf of Mexico (Johnston, 2013). At Sleipner field, with support from the frequently acquired 4D seismic monitors, nearly 16 million tons of CO2 has been stored to the reservoir (Xue et al., 2017). To quantitatively maximize the value of information captured by the 4D surveys, the 4D seismic data has to be assimilated to the reservoir prediction models. This data assimilation procedure is called seismic history matching (SHM), which closes the loop between the observed 4D seismic (and production) and that predicted by reservoir models (Gosselin et al., 2003; Stephen et al., 2006; Yin et al., 2017; Zhang and Leeuwenburgh, 2017). The objective is to quantitatively reduce the uncertainty surrounding reservoir management decisions, by obtaining reliable prediction of reservoir behaviours using history-matched reservoir models. It is believed that the 4D seismic adds spatial constraints to the reservoir simulation models, and thus helps to tackle the problem of non-uniqueness in the ill-posed conventional production history matching (HM).

Because of the valuable information provided by the 4D seismic, frequently repeated seismic monitoring has nowadays become more widely applied in offshore environments through the towed-streamer technology. Table 1 provides a snapshot of the fields that have five or more repeated 4D seismic surveys from the literature. To enhance the data quality and obtain seismic monitors more frequently, seabed permanent reservoir monitoring (PRM) has become popular to provide life-of-field seismic surveys. Table 2 collects a number of the major fields throughout the world that have the PRM system installed. For these fields, the permanently installed acquisition system delivers well resolvable 4D seismic data within a rapid-turn-around processing time. For example, in Ekofisk, it enables excellent 4D seismic data to be obtained every six-months (Grandi et al., 2013). The most recent progress has come from the continuous seismic monitoring technique of “SeisMovie” (Mateeva et al., 2015). With a land buried source and receiver arrays, this PRM system offers time-lapse seismic data that can image subtle reservoir changes on a daily basis. These frequently acquired surveys enable the 4D seismic data to impact the reservoir development decisions promptly regarding infill well drillings, well interventions and production optimization.

However, such frequent 4D seismic acquisition imposes new challenges to the conventional SHM workflows. First, history matching to multiple 4D seismic surveys requires the ability to handle large volumes of seismic data created by the high number of seismic surveys. It can be very computationally demanding and time-consuming if using the traditional history matching workflow that attempts to match each individual time-lapse seismic. Besides, using 4D seismic directly as history matching input data without proper analysis may lead to seismic errors being propagated systematically to the quantitative SHM workflows (Alfonzo et al., 2017). This is because as an ill-posed inverse problem, SHM is sensitive to data errors which means small errors in data the can result in large fluctuations in the prediction (Li, 2017). This problem can further propagate with the increased number of 4D seismic surveys.

In this paper, we propose a framework to efficiently assimilate the frequently acquired 4D seismic data in SHM. A new 4D seismic attribute (named “well2seis”) for history matching is introduced that condenses the many repeated 4D seismic data into a single unitless attribute by correlating them to the reservoir production performances. This not only compresses the big volumes of frequently acquired 4D seismic into a single attribute for efficient assimilation, but also reduces the uncertainty of the 4D seismic observations by summarizing them based on production performances. Morris sensitivity analysis is adapted to investigate the uncertainty parameters in the reservoir model. Once the reservoir uncertainty parameters are confirmed, a well2seis objective function (OF) is constructed to quantify the misfit between the observed well2seis (calculated using observed 4D seismic and production data) and modelled well2seis (calculated using modelled 4D seismic and production performances). In the conventional SHM, because production and seismic data are in different metrics (Chassagne et al., 2016), the weights on seismic and production misfits have to be properly specified when combining them into the OF. But this is avoided when using well2seis OF, as it uses only the well2seis to calculate the misfit, while summarizing information from both 4D seismic and production data. ES-MDA is applied at the end to minimize the well2seis OF by calibrating the uncertainty model parameters identified by the Morris sensitivity analysis. This proposed workflow is tested on a North Sea field reservoir and compared to the conventional production and seismic history matching practices.

Section snippets

Condensing the frequently repeated 4D seismic data

All the reservoir-induced dynamic changes detected by 4D seismic data are caused by fluid extraction or injection activity from the wells. 4D seismic signals therefore cannot be unambiguously interpreted without a clear understanding of the field production and injection behaviours. Considering a reservoir with n repeated time-lapse seismic surveys acquired during the development, a total of N = n*(n-1)/2 4D seismic differences can be generated for all paired combinations of surveys. They will

A North Sea reservoir case

To demonstrate the application of the above proposed data assimilation workflow, we applied it to a reservoir model derived from a real North Sea field with frequently repeated 4D seismic acquisitions. With the geological features and development history kept unchanged as the real field, this reservoir model is generalized to be a prototype reservoir case to investigate the data assimilation problems of history matching. We create a “Truth” case by picking a best history matched reservoir model

Discussions

Our study further develops a cross-disciplinary dimensionless attribute uniting the frequently repeated 4D seismic surveys with the production data to calibrate reservoir models and reduce prediction uncertainty. This well2seis attribute for data assimilation first condenses all the available 4D seismic from multiple surveys into a single attribute, such that it avoids directly history matching to large volumes of seismic data created from the many repeated surveys. This improves the previous

Conclusions

This paper presents a scheme for fast assimilating of frequently repeated 4D seismic surveys that normally contains large volume of measurement data. Instead of using the conventional data assimilation methods that directly history match to the reservoir production history and 4D seismic data, the 4D seismic and production data are first correlated to generate a dimensionless cross-disciplinary attribute, which compresses the information from both the production history and many repeated 4D

Authorship statement

First author: proposed the main workflow, conducted the technical development, drafted and revised this paper. Second author: provided critical insights by implementing the proposed workflow to real field assets, and assisted drafting of the manuscript. Third author: supervised the research project and provided critical insights in drafting and revising the manuscript.

Computer code availability

The code and manual of the Ensemble Reservoir Tool (ERT) used for this research can be obtained freely from: http://ert.nr.no/ert/index.php/The_ERT_code.

Acknowledgments

We thank Equinor for sponsoring this research project. We also want to thank the sponsors of the Edinburgh Time Lapse Project Phase VI (Aker BP, BG, BP, CGG, Chevron, ConocoPhillip, ENI, Equinor, ExxonMobil, Halliburton, Hess, Ikon Science, Maersk, Nexen, Norsar, OMV, Petrobras, RSI, Shell, and TAQA) for supporting this research. The authors would like to specially thank Milana Ayzenberg, Rachares Petvipusit, Matteo Ravasi, Romain Chassagne, Hamed Amini and Mingyi Wong for the technical

References (49)

  • H. Amini

    A Pragmatic Approach to Simulator to Seismic Modelling for 4D Seismic Interpretation

    (2014)
  • M. Ayzenberg et al.

    Accessing a North Sea reservoir connectivity from 4D seismic and production data

  • R. Behrens et al.

    4D seismic monitoring of water influx at bay marchand: the practical use of 4D in an imperfect world

    SPE Reservoir Eval. Eng.

    (2002)
  • R. Brooks et al.
    (1964)
  • J. Caldwell et al.

    Seismic permanent reservoir monitoring (PRM)–A growing market

    First Break

    (2015)
  • S. Campbell et al.

    Optimising value through improved 4D seismic processing on 10 vintages-foinaven-schiehallion-loyal case history

  • F. Campolongo et al.

    An effective screening design for sensitivity analysis of large models

    Environ. Model. Softw

    (2007)
  • R. Chassagne et al.

    An analysis of the seismic history matching objective function

  • C. Daniel

    One-at-a-Time plans

    J. Am. Stat. Assoc.

    (1973)
  • A.A. Emerick et al.

    History matching time-lapse seismic data using the ensemble Kalman filter with multiple data assimilations

    Comput. Geosci.

    (2012)
  • A.A. Emerick

    Deterministic ensemble smoother with multiple data assimilation as an alternative for history-matching seismic data

    Comput. Geosci.

    (2018)
  • M. Eriksrud

    Seabed permanent reservoir monitoring (PRM) – a valid 4D seismic technology for fields in the North Sea

    First Break

    (2014)
  • G. Evensen

    Sequential data assimilation with a nonlinear quasi‐geostrophic model using Monte Carlo methods to forecast error statistics

    J. Geophys. Res.: Oceans

    (1994)
  • C. Geng et al.

    Seismic history matching using a fast-track simulator to seismic proxy

  • Cited by (37)

    • A vector-to-sequence based multilayer recurrent network surrogate model for history matching of large-scale reservoir

      2022, Journal of Petroleum Science and Engineering
      Citation Excerpt :

      The determination of principal component eigenvectors generally adopts the energy criterion or simply uses the first Nr eigenvectors in U. In addition, the sensitivity analysis methods (Nossent et al., 2011; Park et al., 2016; Yin et al., 2019) can also be used to select a subset of U as the Ul. Note that we use the PCA realizations to run the numerical simulation to build the training samples rather than the prior realizations of geostatistical modeling.

    View all citing articles on Scopus
    1

    Now at: Department of Geological Sciences, Stanford University, CA, 94305.

    View full text