The challenge to predict the future is the very reason why we developed sophisticated methods to integrate data, whatever the domain of application. In the context of the subsurface flow characterisation, the 4D seismic data reflecting the reservoir time-lapsed changes as a result of production, requires detailed processing before derived valuable information is incorporated in a subsurface model update. This is a highly uncertain time-consuming process with embedded uncertainty, which is still a challenge in data assimilation procedure. In this study, we propose a new way to the problem of forecasting. A new framework is proposed to extract a proxy model for front prediction directly from time-lapse seismic data. Multiple time-lapsed seismic surveys present the evolution of processes happening in the subsurface. These, therefore can provide information on the propagation of saturation fronts. This paper is based on the simple idea that, different snapshots in time should be able to determine a velocity propagation of fronts and from this velocity, we can then predict their future unknown positions. This work applies a modified version of the Eikonal equation and a modified version of the fast-marching method. This new approach of forecasting a front position is a fast track prediction of fronts evolution. Also, this approach is more strongly based on selected information within the data. To test this approach, we used a water injection scenario from a North Sea reservoir, from which we generated the synthetic seismic data. We compared the results from our algorithm with results from the simulation model, showing that with few data inputs (only two fronts) and a short computing time of 90 s, this method is able to give a good approximation of the waterflood front evolution. It functions as a rapid solution for front prediction based directly on seismic data, irrespective of the shape of the front (isolated blobs), with the only prerequisite being that the sequence of measured fronts is monotonic.
|Journal||Computers and Geosciences|
|Early online date||20 Jan 2020|
|Publication status||Published - Mar 2020|