摘要

Cross-hole electromagnetic ( EM) inversion has been increasingly used to study the variations in rock properties between boreholes, particularly the properties and spatial distribution of the fluids in a subterranean reservoir. In this study, a practical method for simulating and inverting cross-hole EM data to reconstruct a resistivity distribution is presented. The iterative regularized inversion method incorporates a priori information and variable roughness measures, assuming that the medium in the target area is two dimensional and the injected source is a point dipole source. A forward solver for two- and-half-dimensional (2.5D) EM modelling was developed using the finite element method (FEM) for evaluating the forward problem and the 'pseudo-forward' problem needed to construct the sensitivity matrix. The resulting linear system of equations in the forward problem is solved efficiently by the modified incomplete Cholesky biconjugate gradient iterative method. The partial derivatives forming the sensitivity matrix are computed by the adjoint equation based on the reciprocity principle. The regularized inversion scheme, which can be constrained using the available a priori information from well logging, is used to determine the inter-well resistivity distribution from synthetic EM data and the field data acquired in Gudao oil field in east China. The synthetic and field data examples indicate that our scheme is robust and stable in the presence of random noise and therefore can be used for general cross-hole EM interpretation.