Objectives and problems to be solved: The main aim of the project is to develop methodologies for integration of data derived from time lapse seismic surveys into the process of history matching of simulation models of fluid flow in hydrocarbon reservoirs. There are a number of outstanding problems that need to be resolved to achieve this. These include issues related to the interpretation of the seismic data, the petrol-elastic modelling, the differences in scale between model and observations, the uncertainty in the interpreted data and the impact of the new data type on the history matching (optimisation) procedure itself. If the project is successful, reservoir characterization will be improved resulting in more accurate predictions from simulation models. This in turn should lead to more cost effective and more efficient production of hydrocarbons (key action 6.4.1 in the NNE work programme). Description of work: The first task is to assess the impact of incorporating saturation and pressure change data in computer-assisted history-matching based on gradient methods. These new data are distributed throughout the reservoir, in contrast to the usual production data at the wells. The project uses both real and synthetic cases to assess and improve algorithms for computing the covariance matrix of the new data, for parameter selection and for optimisation. This analysis is then extended to matching changes in elastic parameters inferred from the time lapse seismic, by using a petroelastic model within the history-matching process. Comparison of these two approaches is a key step, and includes assessment of the accuracy and reliability of the interpreted data, as well as reconciliation of the different scales of the reservoir models and the seismic data. The project will also examine the application of advanced interpretation techniques, including visualization techniques such as virtual reality, to real time-lapse seismic data. The data used in the project will come from a major oil reservoir in the North Sea and two gas reservoirs in the Adriatic. State-of-the-art petrol-elastic models for these reservoirs, calibrated to available experimental results, will be used. Working with actual field data will enable the developed methodologies to be validated by application to real cases. Software development is an integral part of the project. Research software will be used to test ideas, the successful ones being transferred to commercial code. The goal is to have the chosen methodologies implemented in a commercial history-matching software package, in prototype form, by the end of the project. Expected Results and Exploitation Plans: The first major project deliverable is a proven methodology for incorporating time-lapse seismic data in the history-matching process. There is an immediate benefit to the participating oil companies in using this to exploit their hydrocarbon reservoirs more efficiently. This leads to improve competitively and thus greater security of employment. The environmental impact of hydrocarbon extraction is also reduced. The second major deliverable is a prototype commercial software product. Subsequent commercialisation should lead to profitable sales, both of this unique software and of consulting services. The project contributes to keeping European companies at the forefront of a highly technology-dependent and rapidly evolving industry.
The approach adopted in the project has been to extend existing history-matching software to incorporate seismic-derived data. The software uses gradient-based methods to minimise an objective function, which is a measure of the deviation between observed and simulated data. The new seismic-derived data is distributed throughout the reservoir, in contrast to the usual production data that is located at the wells. The first possible form of this data is as saturation and pressure changes that have occurred within the reservoir between two seismic surveys. The project has used a synthetic case to assess the validity of this approach. This test has proven that the new information can be used efficiently and that its combination with traditional production data can be very effective in constraining the reservoir model. The second possible form of this data, which avoids the difficult step of quantitative identification of reservoir fluid changes, is as the change in the elastic properties of the reservoir between the two surveys. The history-matching process has been extended to matching the change in elastic parameters inferred from the time-lapse seismic, by using a petro-elastic model.
This approach requires a seismic inversion of time-lapse data for all surveys, and a coupling between the fluid flow simulator and the petro-elastic simulator within the optimisation loop. The project has established a general framework to derive elastic impedance from rock and fluid flow properties, which include state-of-the-art petro-elastic models for the reservoirs being studied, calibrated to available experimental results. The key issues with this approach have been examined first through a synthetic case. The incorporation of elastic data within the history-matching process has been successfully achieved for this case, and has proved to be equally effective in constraining the reservoir model when combined with well production data. In both approaches the parameter selection has been extended to incorporate the new data. During the project, alternative formulations of the objective function have been evaluated, especially how the seismic data should be weighted relatively to the production data in a combined objective function. Other issues that have been investigated include: the way of computing the covariance matrix of the new data; the assessment of the accuracy and reliability of the interpreted data; reconciliation of the different scales of the reservoir models and the seismic data; improvement of the optimisation algorithms. Working with actual field data has enabled the developed methodologies to be validated by application to real cases.
The project has examined the application of advanced interpretation techniques, including visualisation techniques, to real time-lapse seismic data. The data used in the project comes from a major oil reservoir in the North Sea and two gas reservoirs in the Adriatic. Reservoir simulation models of these reservoirs have been created and matched to conventional production data. This provided the baseline for comparison with models matched to both seismic-derived and production data. The application of the proposed workflow to three real field cases has stressed the necessity of accurate estimates of variances and correlations for the time-lapse seismic data. The sensitivity analysis to select the most effective matching parameters has been adapted to the new kind of data. The overall methodology has proven to be useful for improving the reservoir characterisation of the cases under study, but other application cases will be necessary in the future to develop further these novel techniques. The project has used research software to test ideas, with the aim of transferring the successful ones to commercial code. A commercial reservoir simulator has been extended to generate gradients of standard simulation properties (e.g. pressure, saturation, density) on simulation grid-cells, and a petro-elastic model has been incorporated in order that the simulator can calculate values of seismic-derived properties in the acoustic domain (e.g. acoustic impedance). Also, a commercial computer-aided history-matching tool has also been modified to use these gradients to perform matching of seismic-derived data at the same time as production data (e.g. pressures and flow rates in wells).
Funding SchemeCSC - Cost-sharing contracts
20097 San Donato Milanese
OX14 1DZ Abingdon