The aim of this paper is to propose an approximate technique for the reconstruction of magnetic-field distribution in the proximity of unknown sources. The method is based on two nested optimization algorithms-the external one is responsible for the definition of the geometry of the equivalent-source system, while an internal procedure based on least-square minimization is used to identify the currents of the sources. The goal is to minimize the deviation between the reconstructed and actual field. It is shown that under magnetic quasi stationary conditions, the proposed approach is able to reconstruct the field distribution of a complex 3-D environment as medium-voltage/low-voltage substations. The main limitation is the availability of magnitude and phase of the measured magnetic field. This information is not usually provided by standard probes but can be obtained by a proper measurement setup. The proposed technique allows accurate reconstruction of the magnetic field even in the case of many sources without requiring their accurate geometrical description and is suitable for the evaluation of stray fields and the design of magnetic shielding.