Foothill Structure Inversion Using Multi-Objective Evolutionary Algorithm

Singh, Vijay Pratap¹, Michel Léger¹, Marc Schoenauer² (1) Institut Francais du Petrole, Rueil-Malmaison, France (2) INRIA Futurs, Orsay, France

Constraining unexposed or poorly resolved subsurface fault and layer geometry from observed data is extremely important for evaluation of hydrocarbon potential in mountain­ous regions, and also for the assessment of seismic hazard. Construction of kinematic mod­els for a mountain front is generally very complicated, ambiguous and a lengthy process; in particular it is difficult to come up with a single solution that fits to the geologic observa­tions.

We are proposing a practical approach which will enable us to get the true solution and simultaneously automate the simulation process. For the first time, a multi-objective evolu­tionary algorithm (MOEA) has been successfully used for geological optimisation, as far as our knowledge of literature goes. MOEA ability to search in complex spaces is of utmost use in these situations. In this approach, populations of kinematic models are generated ran­domly and optimised using MOEA on the basis of available information and field data. The objective functions of MOEA are L2 norms about the dip of faults and layers, and the fault location. The surface and well data is used for this inversion.

A set of optimal solutions are obtained in a single run. These solutions are well matched with the observed data. The combination of higher order information and the knowledge of an experienced observer will increase the accuracy of detection by many fold.

Later on, we will synthesize this geological approach with seismic velocity model opti­misation.