Abstract: Accurate determination of aquifer hydrological parameters, such as permeability coefficient, is essential for effective mine water hazard prevention and control. However, traditional inversion methods such as the fitting curve method and graphical method exhibit shortcomings in computational speed and accuracy. To enhance the reliability of aquifer parameter inversion calculations, this study proposed a novel permeability coefficient inversion model, the Adaptive Differential Hybrid Butterfly Particle Algorithm (ADHBPA), specifically tailored to the characteristics of hydrogeological parameters. The model incorporates Latin hypercube sampling, a hyperbolic cosine adaptive function, differential mutation strategy, and dimension-wise variation strategy. The model effectively addressed the spatial heterogeneity and temporal dynamics inherent in hydrogeological parameter inversion, thereby improving the balance between global exploration and local exploitation. Using the pumping test data from 24 boreholes in the Banji mining area, the ADHBPA model achieved a maximum inversion error of 0.93 m and an average error rate of just 0.15%. In contrast, conventional algorithms produced average error rates ranging from 30% to 50%. These results highlight the algorithm's strong capability in avoiding local optima and performing high-precision parameter inversion, even under data-scarce conditions. The proposed algorithm provides efficient and reliable technical support for mine water hazard risk assessment and water control planning.