Two-dimensional Inversion Of Marine Controlled-source Electromagnetics With Undulating Seafloor Topography Based On Unstructured Finite Element Method

As one of the main electromagnetic exploration methods,marine controlled source electromagnetic(MCSEM)is widely used in offshore oil exploration with its advantages of high efficiency,low cost and high abnormal resolution.Marine electromagnetic exploration is a newfashioned geophysical exploration method rising from the 1970 s.It plays an important role in the exploration of metal and non-metal ore,geological structure investigation and oil-gas resources development.In this paper,based on the adaptive finite element forward modeling with unstructured triangular mesh,the two-dimensional Occam inversion algorithm of marine controlled source electromagnetic(MCSEM)data with undulating seafloor topography is studied.First of all,this paper uses the unstructured mesh to divide the model area.As the elements of the unstructured mesh are not only be qualitatively controlled,but also be easily local refined,the accuracy and efficiency of finite element solution with two-dimensional unstructured mesh are greatly improved in simulating the complex geometric models.Compared with structured grids,the unstructured grids are more flexibly and efficient.Unstructured grids can accurately accommodate the structures with different scales and complex topography,which are not designed artificially.Consequently,the positions and orientations of the transmitters and receivers placed on the submarine slopes can accurately be simulated with the structured grids,and the accurate solutions of the electromagnetic fields can be obtained.In this paper,the two-dimensional forward modeling of MCSEM based on adaptive finite element with unstructured mesh is analyzed.With numerical examples,the electromagnetic characteristics of MCSEM and mesh refinement features are analyzed.The numerical examples show that the accuracy of the forward simulation of the MCSEM fields can be effectively improved by using the adaptive finite element method,which adjusts the mesh automatically according to a posteriori error estimation of the FE solutions.A more refined finite element grid is necessary for more accurate response,but considering that the amount of inversion parameters will also be increased largely,hence,the unstructured double grids system is adopted in this paper: one is the inversion grid,usually a coarse grid,and the other is the forward modeling grid,usually is obtained from the refinement of the inversion grid,in this paper,the adaptive finite element method is applied to solve the problem,which ensures the accuracy of forward modeling as well as the efficiency of inversion.All in all,this unstructured double grids system is more significant for both the forward modeling precision and inversion efficiency than the traditional single grid system.MPI parallel computing is conducted to speed up the forward computation of MCSEM fields with multiple sources and multiple frequencies.The speed of computing is compared between the serial computing and parallel computing with multiple processes.The efficiency of the parallel algorithm is verified by a model test.MPI parallel computing speeds up the forward and inverse processes to a certain extent in considering the performance of the computer.With the number of parallel processes increased,the speed up ratio is increased,but the parallel efficiency is decreased.In the inversion,the fast Occam inversion algorithm is adopted,which is improved from the conventional Occam algorithm.In the inversion process,if at any given step of the minimum search a model is found with a sufficient misfit reduction compared to the staring model's misfit,it is accepted and the iteration is terminated,normally a RMS misfit reduction threshold of about 15 percent has been found to work well;otherwise the complete minimum search is carried out as normally done for the regular Occam algorithm.The fast Occam inversion can not only reduce the amount of model search,but also reduce the number of forward calculation,which makes it a hybrid that balances the speed of fixed algorithms and the robustness of the Occam's dynamically determined .The inversion examples of the two-dimensional marine oil and gas reservoir model are carried out with different measure devices.The results show that the inversion results can converge stably to the vicinity of the true model.Through the inversion of terrain model,the effect of twodimensional inversion of MCSEM data incorporating with undulating terrain is analyzed,which provides a reference for the accurate interpretation of MCSEM data.