Controller placement in Software-Defined Networking (SDN) plays a pivotal role in determining latency, energy efficiency, and network adaptability. However, traditional static optimization methods often fail to meet the demands of dynamic and large-scale environments. This paper presents a dual-stage hybrid optimization framework that integrates Non-Dominated Sorting Genetic Algorithm III (NSGA-III) with Proximal Policy Optimization (PPO) to achieve adaptive and energy-efficient controller deployment. NSGA-III is employed to generate Pareto-optimal solutions considering latency, energy consumption, and load balancing, while PPO continuously refines these placements based on real-time network states. Furthermore, an energy-aware traffic-routing mechanism is introduced to dynamically reroute flows according to link utilization and energy metrics. Evaluations on the GEANT topology show that the hybrid model reduces latency by up to 18.5%, lowers energy consumption by 21.7%, and improves load distribution by 16.2% compared to baseline methods. The results underscore the effectiveness of combining multi-objective evolutionary optimization with reinforcement learning for robust and scalable SDN controller placement.