2D Finite Element Modeling And Inversion For Marine Controlled-source Electromagnetic Fields

In the21st century, oil exploration has been turned from land to ocean.Although three-dimensional seismic survey has been one of important means， whichcan provide submarine stratigraphic structure for the ocean drilling, oil companies arenot satisfied with getting the accurate knowledge of submarine structures, at the sametime, they also want to know the properties of the fluid within the structure in order toimprove the drilling success. Resistivity is sensitive to lithological variations and ithas became one of most important parameters for reservoir interpretation, such asresistivity logging data interpretation. Marine controlled source electromagnetics is ageophysical method for oil and gas resources exploration, and it can detect theanomalies associated with hydrocarbon reservoir effectively utilizing the resistivitydifferences. The development of marine controlled source electromagnetic equipmentand methodology have been obtained with great successes together with its numericalforward modeling and inversion interpretation, while studies in these areas are laggingbehind in China. This is the main purpose of the thesis. This thesis discusseed themain theoretical foundation and key techniques for modeling and inversion of marinecontrolled source electromagnetic data, includeing:(1) This work proposed expressions of electromagnetic fields excited by electricdipole source in isotropic/anisotropic layered medium using vector potentials startingfrom the Maxwell’s equations. An arbitrarily orientated source is decomposed intothree equivalent components along three Cartesian coordinates, and its field responsescomputed by fast Hankel transforms are given by superposition of those of threeequivalent ones, respectively. Electromagnetic fields excited by a finite dipole arecomputed using the integration or the linear superposition method. In short offset area, the finite dipole cannot be taken as a point dipole. Usually an integration method isapplied to compute the field responses in FORTRAN code and verify its accuracy andprecision.(2) Based on the three-layer background model with air-water-rock, this thesisanalyses the relationship between the amplitude of the electromagnetic field responsesand the seafloor rock resistivity, then a method to calculate the equivalent resistivityusing marine oil-gas model is proposed. This method provides qualitativeinterpretation for marine controlled source electromagnetic data.(3) This dissertation gives expressions of layered isotropic electromagnetic fieldsand deduces partial derivatives of field responses with respect to conductivity. Alsoone-dimensional inversion of marine controlled source electromagnetic field isdeveloped using Gauss-Newton method.(4) Adaptive finite element method is applied for solving the boundary conditionproblem of marine CSEM for both total field and primary-secondary field. Theprimary-secondary approach can yield stable and accurate solutions by computing theprimary EM field and the secondary EM field, respectively. Some difficulties in2Dmarine CSEM modeling were analyzed in this thesis:1) different pseudo deltafunctions are used to distribute dipole current source, and an optimal selection isobtained by analyzing the performance of each discrete method;2) For solving thelinear equation system formed by CSEM finite-element modeling, I compared theperformance of several direct solvers;3) adaptive finite-element methods canautomatically improve mesh generation, offering reliable solutions at a reasonablecomputational cost and simulate complex geoelectric models;4) as CSEM2Dforward modeling is extremely time-consuming, parallel computation for thesimulation is performed.(5)2D CSEM inversion is developed based on our forward modeling algorithm.Several key factors, such as the regularization factor, model constraints,, are analyzed.Three different methods are employed to derive the Jacobian, and their computationalefficiency is analyzed. A Gauss-Newton approach is used with the conjugate gradientmethod for linear search. For limited memory, only the product of Jacobian matrix and vector is calculated to avoid direct calculating and storing the Jacobian matrix.The inversion result of synthetic data shows that the algorithm can recover the seabedoil and gas layer well.