A Phaseless Source Reconstruction Method Based on a Multipopulation Differential Evolution Algorithm With Divided Strategies and Dynamic Adjustment

This article proposes a novel phaseless electromagnetic source reconstruction method, referred to as adaptive subpopulation multistrategy differential evolution. The method integrates a differential evolution (DE) algorithm with a multipopulation partitioning strategy and degrees of freedom (DOF) estimation. DOF estimation is a process that quantifies field distribution complexity by analyzing spatial change rates. This quantitative parameter is used to guide mathematical modeling, thereby eliminating the arbitrary selection of the number of dipoles. The method also employs joint generalized cross-validation (GCV) Tikhonov regularization. The joint GCV Tikhonov regularization enables the automatic, data-driven selection of the regularization parameter, which eliminates the subjective and inefficient nature of manual trial-and-error. This approach effectively constructs equivalent dipoles that represent actual electromagnetic interference sources and delivers accurate near–field reconstruction results.