To address the problem of local optima caused by the complexity of the objective function in parameter inversion for groundwater numerical modeling, an improved quantum genetic algorithm (IQGA) was proposed. Based on niche technology and elitist preservation strategy, a simulated binary crossover method varying with the number of iterations to replace the traditional rotation gate strategy was introduced. Furthermore, the improved algorithm was combined with the dynamic time warping (DTW) technique. Population diversity was maintained through the niche mechanism, while global search capability was enhanced by dynamically updating the population using an elite solution set and a crossover strategy that evolved over iterations. Additionally, DTW distance was employed to quantify the morphological similarity between simulated values and time-series observational data, facilitating the selection of globally optimal parameter combinations. The performance of IQGA was evaluated through benchmark functions, a relatively ideal case study (Case 1), and a real-world complex case (Case 2), and was compared with traditional genetic algorithm (GA), differential evolution (DE), particle swarm optimization (PSO), and basic quantum genetic algorithm (QGA). Experimental results show that during the benchmark function tests, the minimum, maximum, and average errors obtained by IQGA are all superior to those of the other algorithms. Convergence stability is significantly improved, and IQGA consistently outperforms the other algorithms in both the ideal and complex case studies. These results indicate that IQGA, through a synergistic optimization mechanism, effectively avoid the trap of local optima and provide a reliable approach for high-precision parameter identification in complex groundwater models.