Abstract: To address the high computational cost of Monte Carlo Simulation (MCS), the insufficient convergence efficiency of the Active Kriging–Monte Carlo Simulation (AK-MCS) method in high-dimensional problems, and the limited local search capability of the traditional U-learning function, this paper proposes a structural reliability analysis method based on an Adaptive Model Ensemble Kriging framework, termed AME-AK-MCS-L. First, a high-density uniform initial sampling strategy based on a space-refilling mechanism is employed. By incorporating local void detection and regional refilling procedures, the overall coverage of the design space and the accuracy of the initial Kriging surrogate model are significantly improved. Second, a novel L-learning function is developed by introducing the probability of prediction error, sample-space distance, and a dynamic exploration factor into the conventional U-learning function, thereby enhancing both global exploration and local exploitation capabilities. Furthermore, an adaptive ensemble surrogate model is established, in which the prediction results of multiple sub-models are integrated through dynamic weighting to improve the accuracy and robustness of reliability analysis. The proposed method is validated using two benchmark examples: a low-dimensional series system with four failure domains and a high-dimensional six-dimensional nonlinear oscillatory system. Comparative studies with AK-MCS-U, AK-MCS-L, and AME-AK-MCS-U demonstrate the effectiveness and superiority of the proposed approach. The method provides an efficient and reliable solution for structural reliability assessment of high-dimensional complex systems.