<p>Accurate modeling of charge distribution plays a vital role in molecular simulations, electrostatic energy evaluation, and mechanistic analysis. The atom-bond electronegativity equalization method (ABEEM) provides a physically interpretable framework for computing atomic and electronic site charges by partitioning molecular space into atoms, bonds, and lone-pair regions. However, conventional ABEEM parameterization relies heavily on manual tuning, limiting its adaptability and predictive accuracy. In this work, an automated parameter learning strategy for ABEEM was proposed, guided by intelligent optimization algorithms and formulated within a goal programming framework. The framework systematically calibrates the key parameters of multiple types of charge sites. A chemically diverse training set including proteins, lipids, and nucleotides was constructed, and a dual-level objective function was designed to improve accuracy at both site and atomic levels. This approach significantly enhances the predictive performance and consistency of the ABEEM model across complex biomolecular systems. It also eliminates human bias and provides a scalable and generalizable pathway for force field development.</p>