Objective The inversion of the horizontal electric field component in the grounded-source short offset transient electromagnetic (SOTEM) method using traditional algorithms is prone to fall into local extrema. To address this challenge, this study proposed an improved particle swarm optimization (PSO) algorithm that integrates the center of gravity reverse learning strategy.

Methods Based on the center of gravity reverse learning strategy, the improved PSO algorithm can dynamically adjust learning factors and the value of the adaptive inertia weight, thus improving the global search capability and convergence efficiency effectively. The performance of the improved PSO algorithm was verified using the typical three-, five-, and seven-layered geoelectric models constructed in this study.

Results and Conclusions The results of this study indicate that for the five- and seven-layered geoelectric models, the damped least squares method yielded average inversion errors of 0.34% and 4.68%, respectively, while the improved PSO algorithm yielded average inversion errors of 0.21% and 0.87%, respectively. This suggests that the improved PSO algorithm significantly improved the identification accuracy of complex geoelectric structures. Under the conditions of multi-layer (≥5) initial inversion intervals and wide search intervals, the improved PSO algorithm yielded average inversion errors of less than 5% for both three- and five-layered geoelectric models, substantiating its effectiveness. Inversion was conducted for the measured data from a certain mining area using the damped least squares method and the improved PSO algorithm. The inversion results demonstrate that the improved PSO algorithm outperformed the damped least squares method, with the inversion results of the improved PSO algorithm agreeing well with the electrical structure of the known ore body. The results of this study will provide theoretical support for improving the resolution of SOTEM in mineral exploration.