History Matching Using Time-lapse Seismic (HUTS)

Abstract

Abstract This paper describes the results of a two-year EC-sponsored project which uses new information provided by repeated seismic acquisitions (4D seismic data) jointly with production data in an extended, efficient and consistent history matching process. This process involves a simultaneous minimisation of the mismatch between all types of measured and simulated data. A gradient-based technique has been developed and tested both in a prototype and in commercial computer-aided history matching software. We show results on real cases, located in the North Sea and the Adriatic Sea, and discuss key issues of such seismic history matching. Most applications of time-lapse seismic to date have been qualitative or semi-quantitative. We propose a quantitative work flow. The seismic contribution in the objective function is defined in terms of elastic parameter variations within the reservoir and the data have been properly scaled using an estimate of seismic uncertainty (covariance matrix). The "observed" values are obtained by inversion of the seismic signal. For the "modelled" values, the flow simulator is coupled with a petro-elastic model to convert simulated fluid and static rock properties into simulated elastic properties. The techniques described in this paper allow us to reconcile production history matched models with 4D information, and to reduce the uncertainty in reservoir properties, which haven't a real impact on the well history, but which significantly drive future behaviour of the field. This is a further step towards the necessary integration of available data for better predictive simulations. Focusing on quantitative combined with qualitative use of data enhances the multidisciplinary approach. Introduction The use of time-lapse, or 4D seismic data in reservoir management, characterisation and monitoring is steadily increasing as the interpretations get more reliable. Inverted seismic data has proven to be very valuable for locating remaining oil and plan infill drilling.1,2 Better and more reliable 4D data also triggers the need for a tool, which can help the petroleum engineers to condition the reservoir models to this data. Several authors have considered the problem of using 4D data in the process of history matching reservoir simulation models, but very few have described a complete software system which can handle the integration of both production data and seismic data in a computer aided history matching loop. Also most papers either use synthetic data3–8 or include the data in a qualitative way9–14. Landa and Horne15 presented a sensitivity study looking at the relative influence of the various types of data in the reservoir characterisation process. Huang et al.16–17have presented a quantitative approach where the misfit between 4D real and synthetic amplitude from reservoir simulations is minimised using a stochastic search procedure. The misfit function also includes production data, but there is a lack of documentation about the exact definitions and algorithms used. The procedure was applied to a Gulf of Mexico dry gas field and improved the reliability of model predictions. Waggoner et al.18 have used a similar approach to a simple gas condensate reservoir. Here the similarity between acoustic impedance variations from 4D data and impedance calculated by a numerical simulator was maximised using a greedy global optimisation algorithm. These approaches, however, apply a global optimisation routine, which requires hundreds and even thousands of simulation runs to obtain a match.