Fog computing provides decentralized resources closer to end-users, reducing latency and improving responsiveness in Internet of Things (IoT) applications. However, energy consumption remains a significant challenge in these diverse environments. This study offers a comparative evaluation of three nature-inspired metaheuristics ‒ Genetic Algorithm (GA), Firework Algorithm (FWA), and Grey Wolf Optimizer (GWO) ‒ for energyefficient application scheduling in fog infrastructures. We use power-aware fitness functions and extensive parameter tuning to balance exploration and exploitation within a fixed computational budget. With over 100 independent runs, GA achieves the lowest final power consumption, averaging 12.1% less than FWA and 25.4% less than GWO. Although FWA reaches its optimal solution approximately 18% earlier than GA, GA maintains better energy efficiency. GWO converges most slowly and produces less efficient solutions despite sustaining higher population diversity. These results highlight the trade-offs between efficiency, conv