24
4d seismic and history matching - activities at the IOR center [email protected]

4d seismic and history matching - activities at the IOR center · Data assimilation using 4-D seismic data TNO’s approach • Implementation of TNO’s history matching workflow

  • Upload
    hamien

  • View
    217

  • Download
    0

Embed Size (px)

Citation preview

4d seismic and history matching- activities at the IOR center

[email protected]

Reservoir characterization

• Sources of information– Geological interpretation– Seismic– Well-logs & cores– Production data– 4-D seismic– Tracer data– Gravity data– …

• Aim: Build a reservoir simulation model for forecasting

Ensemble Kalman filter (EnKF)

• Data assimilation technique introduced in 1994 by Geir Evensen for oceanographic applications

• Data assimilation: Incorporating observational data into a computational model of a physical system

• Today numerous applications of EnKF, e.g.– Atmospheric & meterological systems– Water resources & groundwater flow– Oceanography– Environmental & ecological systems– Petroleum engineering– …

Ensemble Kalman filter for reservoir characterization

• First publication in 2002 (Nævdal et al., SPE 75235)

• Started to work on field cases from 2004• Results from several field cases are

published, including – Raniolo et al., 2013 (IOR 2013)– Chen & Oliver, 2014, SPE-164902-PA– Bianco et al., 2007, SPE 107161-MS

• In tool set of several companies• Free and commercial software available

Ensemble based methods

26 January 2018

› Use an ensemble of models to capture significant uncertainties

› The models can be different/uncertain in millions of parameters

› Update all the uncertain parameters in all the models simultaneously in one update step. (Based on statistics computed from the ensemble)

1)()(

−+=

−+=

RHPHPHKHXdKXX

TT

ffa

Aim at IOR center: 4D seismic history matching for field case

• Available data from Norne field (open data set for research)– History matched with production data using ensemble

smoother• Chen & Oliver, 2014, SPE-164902-PA

• Current aim: History match with 4D seismic data– IRIS: Impedances– TNO: Changes in saturation and pressure fronts

• Working on data set from Ekofisk– Interpretation of changes in pressure, saturation, porosity

• In addition: Theoretical work to get better understanding of uncertainty in 4D seismic inversion

Research challenge: Include 4-D seismic data

• Huge number of measurements is complicating the update schemes

• Needs more modelling– Rock physics model– Effects from overburden– Seismic forward modelling or inversion

• Choice of seismic data:– Traces, AVO, impedances,…

• Noise quantification of seismic data?

Case study in the released Norne data set

Previous Norne study at IRIS:History matched the following parameters by production data only:

Current Norne study started at IRIS:Include both production data and seismic data in the HM loop. Seismic data: Inverted AVO data (acoustic impedance).

Steps on the way to a history matched Norne model

• Inversion:– AVO data -> Acoustic impedances – …. -> saturation and pressure fronts (TNO)

• Find suitable rock physics model– Porosity, saturations -> Acoustic

impedances• Define initial ensemble

– Evaluate geological uncertainties– Parameterization

• Challenge in handling large set of measurements

Data assimilation using 4-D seismic dataTNO’s approach

• Implementation of TNO’s history matching workflow on a field case (Norne field).

• PostDoc: Yanhui Zhang, TNO• Literature:

– Trani et al., SPE J. (2012): Arrival time of saturation fronts (need to run ahead of observation time)

– Leeuwenburgh & Arts, Comp. Geo. (2014): Calculate distance between simulated and observed fronts (Cartesian grids)

– Zhang & Leeuwenburgh, ECMOR XV (2016) + Comp. Geo. (2017): Extension of above approach, including adaption to corner-point grid. Synthetic study on the Norne field.

Estimation of reservoir parameter changes using time-lapse seismic data for compacting reservoir

› Objectives• Better usage of time-lapse seismic data (PRM data) to monitor dynamic reservoir parameter changes• Improvement of existing methods to estimate more realistic changes in dynamic parameters • Quantification of associated uncertainties in the estimates

› Methods

• Utilize time-lapse AVO• Deterministic approach:

o Extension of Landrø’s method (2001)o Demonstrated on Synthetic data for a

compacting field example (Bhakta, 2015)• Bayesian approach:

o Extension of Buland and Omre (2003) and Grana and Mukherji(2014) methodology for compacting reservoir scenario

o Better quantification of uncertainties in the estimates

› Benefits• Improve understanding of the remaining hydrocarbon compartments, infill drilling • Taking full advantage of high repeatability of PRM data, thus reducing uncertainties in the estimates• Support better reservoir characterization and management

(Courtesy, Buizard et al., 2013)

4D seismic for compacting reservoirs

• Tuhin Bhakta, IRIS (Post Doc)1. “Better estimation of pressure-saturation changes from

time-lapse PP-AVO data by using non-linear optimization method.” SEG Annual Meeting, 2015.

2. “Estimation of Pressure-Saturation and Porosity Fields from Seismic AVA Data Using an Ensemble Based Method.” IOR Norway, 2017 (with Luo, Nævdal).

3. “Estimation of pressure-saturation and porosity fields from seismic acoustic impedance data using an ensemble based method”, SEG Annual Meeting, 2017 (with Luo, Lorentzen, Nævdal).

4. Working on field case: Ekofisk, we have received the data form ConocoPhillips. Done synthetic studies testing methodology (similar to no. 3).

Estimation of pressure, saturation and porosity from seismic AVA data

• Bhakta, Luo, Nævdal, presented at IOR Norway• Workflow:

Estimation of pressure, saturation and porosity from seismic AVA data

• Bhakta, Luo, Nævdal, presented at IOR Norway• Workflow:

Brugge model

Grid geometry of Brugge field

Results on Brugge modelLayer 2 of 9

Estimation using seismic impedance data

• Presented at SEG 2017• Bhakta, Luo, Lorentzen, Nævdal

Results on synthetic Nornestudy – Layer 2 of 22

Uncertainty assessment

Bayesian estimation of reservoir properties – effects of uncertainty quantification of 4D seismic data

• Eikrem, Nævdal, Jakobsen & Chen

• Comp. Geosciences 2016• Compare different

uncertainty quantifications of 4D seismic data

• Simplistic quantification based on estimating uncertainty of seismic data in unflooded part gives worse results

• Small case study, simplified seismic modelling

Standard deviation of estimated porosity

Proper uncertainty quantification

Simplistic uncertainty quantification

Bayesian Inversion of Time-lapse Seismic Waveform Data Using an

Integral Equation Method• Eikrem, Jakobsen, Nævdal, 2017• Estimating change in velocity between two

surveys, with uncertainty analysis• IOR Norway: Simple 2D model• Recent work: Marmousi

– Baseline + monitor survey– Double difference– SNR: 4

Results on Marmousi modelVelocity field

Estimating difference in velocity

Summary

• Working towards ensemble based history matching including 4D seismic data

• Ensemble based approach opens up for production optimization incorporating model uncertainty

• Also considering compacting reservoirs