This study presents the SHGBA framework, integrating a modified Bat Algorithm (BA) with the Second Horizontal Gradient (SHG) technique to enhance geothermal reservoir characterization through self-potential (SP) data inversion. The SHGBA approach leverages SHG’s ability to suppress regional effects and BA’s global search to precisely estimate subsurface parameters like amplitude coefficient (K), depth (z), location (xo), shape factor (q), and polarization angle (Ɵ). Unlike conventional SP inversion approaches, SHGBA refines SHG-derived anomalies using BA’s global optimization, reducing noise and improving accuracy in geothermal modeling with minimal reliance on prior assumptions. Tested on synthetic models under noise-free and noisy (5% and 10%) conditions, as well as multi-source models with overlapping anomalies, SHGBA consistently recovered parameters, proving its robustness. When applied SP data from Hi’iaka dike surveys (1973–2012) at Kilauea Volcano, Hawaii, SHGBA successively characterized basaltic dike intrusion parameters and tracked temporal thermal changes. Unlike traditional techniques prone to local minima and reliant extensive prior assumptions, SHGBA’s population-based optimization thoroughly explores parameter space, excelling in nonlinear and multi-parameters SP inversion. Requiring minimal prior constraints, SHGBA is well-suited for geothermal exploration, volcanic monitoring, and reservoir management, enhancing drilling efficiency, monitoring, and sustainable geothermal energy production.