In the NISQ era, quantum annealing (QA) shows strong potential for large-scale combinatorial optimization but faces challenges from hardware constraints and penalty parameter selection. This dissertation addresses these issues by proposing adaptive penalty optimization and a scalable QA framework, applied to wireless communication and port logistics. For multi-UAV networks, berth allocation, and quay crane allocation, the proposed methods leverage mathematical reformulation, reinforcement learning, and hybrid quantum?classical integration to improve solution quality and scalability. Experiments demonstrate that these approaches achieve near-optimal solutions faster than classical baselines, advancing the practical applicability of QA in complex real-world environments.