地面频率域电磁法三维正反演成为大规模电磁数据精细化和定量解释的关键。然而受多类型场源、复杂起伏地形、正演模拟的速度和精度以及三维反演的适用性等问题的困扰，三维反演仍难于在实际勘探中获得广泛应用。为此，本项目拟采用有限体积法求解Lorenz规范条件下耦合势对称方程，利用坐标变换技术模拟起伏地形或地电界面，使用场源等效技术离散不同类型场源，引入聚集多重网格、区域分解以及并行求解技术提升正演计算速度，实现任意场源的地面频率域电磁法三维快速正演。另外，建立不同数据集的雅克比矩阵的统一表达式，采用一种“全局”优化算法—截断高斯牛顿法及多种经典反演方法(非线性共轭梯度、高斯牛顿法、L-BFGS等)实现带地形的地面频率域电磁法三维并行反演。本项目期望以此建立地面频率域电磁三维正反演统一平台，实现不同类型场源、不同数据类型的地面频率电磁三维正演模拟和反演解释，提高我国地面频率域电磁勘探三维精细化解释水平。

Three-dimensional modeling and inversion of ground frequency-domain electromagnetic (FDEM) methods become the key issue in detailed and quantitative interpretation of large-scale electromagnetic data. Due to the singularity of the electromagnetic sources, the speed and accuracy of the forward modeling, the applicability and computing efficiency of the inversion, three-dimensional inversion have very limited application in practical exploration. To solve these problems, we propose to adopt the finite-volume (FV) method to solve the coupling equations in Lorenz gauge. The complicated topography or interface are simulated by hexahedral grids and coordinate transformation. The discretization technology for arbitrary transmitting sources are introduced to represent different kinds of electromagnetic sources. By using the aggregation multigrid solver, domain decomposition technique and parallel computation together, we can significantly improve the speed of forward calculation, thus carry out fast three-dimensional modeling for different kinds of FDEM methods. Meanwhile, a general formula for the Jacob matrix for various types of electromagnetic data is established, which serve as the kernel in the parallel computation of the sensitivity or gradient of the object function. To implement the three-dimensional parallel inversion for electromagnetic data effected by topography, the “global” optimization algorithm – truncated Newton method and other gradient based inversion methods (Nonlinear conjugate gradient, Gauss-Newton, L-BFGS) will be utilized to minimize the object function. We extremely expect to build and implement a unified platform for FDEM modeling and inversion, with which different electromagnetic sources and data can be easily solved. This research will improve the technology and level of detailed processing and interpretation for FDEM data definitely.