The artificial honey bee rearing optimization algorithm for hydraulic parameter optimization in multi-aquifer systems

Hydrological model calibration is crucial for seepage predictions, with parameter estimation a key challenge. Machine learning techniques have advanced for optimization in industries. Inspired by honey bee genetic manipulation, the Artificial Honey Bee Rearing Optimization (AHBRO) algorithm was tested on benchmark problems. Combined with the finite difference method, AHBRO was validated through pumping tests in Tianjin, China. Results show in low-dimensional spaces, AHBRO(WOA) outperforms models like EGA. For the Goldstein-Price function, AHBRO(WOA) achieved a fitness of 3.0000021, with a standard deviation of 1.15 × 10−5, better than EGA (fitness 3.0000145, standard deviation 4.36 × 10−4). In most high-dimensional spaces, BP-ANN improved AHBRO’s performance. In Tianjin tests, the estimated drawdown matched the measured values, showing excellent estimation ability with fewer function evaluations. Compared to EHBMO and EGA, AHBRO achieved results closer to the optimum, with an average NSE of 0.98. AHBRO offers an efficient, accurate solution for hydrological optimization, addressing convergence issues, with significant implications.