This paper investigates the challenges of the multi-UAV system flying in formation in realistic cluttered environments. An adaptable scaling formation method is proposed based on nonlinear model predictive control (NMPC). A formation cost term constructed by using a simple normalized Laplacian matrix promotes UAVs to form a formation with scaling property, which improves the flexibility of obstacle avoidance. A cohesive cost term is designed to prevent the formation from scaling infinitely. A velocity cost term makes the velocity of the UAV controllable, and prompts adaptive scaling of the formation to fit the environment. The energy consumption during flight is optimized using a control cost term. Finally, the effectiveness of the proposed method is verified through simulation.