A Hybrid Global-Local Optimization Algorithm Combining Luus-Jaakola and Levenberg-Marquardt for Self-Potential Inversion

The inversion of self-potential (SP) data to determine parameters of geological models, such as a 2D inclined sheet, faces stability and accuracy challenges due to its non-linear and non-unique nature. This study begins with a parameter sensitivity analysis, revealing that depth and dip angle significantly influence the SP response more than horizontal position and amplitude coefficient. To address these challenges, we developed a hybrid inversion algorithm that integrates the Lous-Jakovac (LJ) global search for initial estimates with the Levenberg-Marquardt (LM) local optimization for refinement. This combination has been tested on both noise-free and noisy synthetic data, as well as real field data, aiming to improve the accuracy of geometric parameters and enhance the stability of the inversion for the inclined sheet model. The results indicate that the hybrid method significantly reduces parameter estimation errors and offers a better fit to the observed data compared to the standalone LJ and LM algorithms, as well as other methods. This approach demonstrates high stability against noise and accurately delineates the ore body's location.