Abstract: Tandem wings share lift and reduce induced drag with the front and back wings, which is a benefit. However, the complex wake interference makes the flow field unstable, which stops the improvement of aerodynamic performance even further. An active jet control method based on transfer deep reinforcement learning is suggested in this paper to get around these issues. To teach the smart agent, the Proximal Policy Optimisation (PPO) method is used. The goal is to stabilise lift and lower drag. This is done by changing the jet strength on the wing surface. For the training, the Reynolds number is set to 1000, the distance between the upper and lower wings is h = 0.5c (aerofoil chord length), and the distance between the front and back wings is d = 2c. The approach was then moved to four different layouts: (a) h = 0.5c, d = 3c; (b) h = 0.5c, d = 4c; (c) h = −0.5c, d = 2c; (d) h = 0c, d = 2c. This was done to see how well it could be used in other situations and how stable it was. The findings show that the lift-to-drag ratio of the front wing increased by 22.89% and that of the rear wing increased by 5.37%. During migration, the lift-to-drag ratio of the front wing increased by 17.27%, 18.03%, 19.35%, and 31.64%, and that of the rear wing increased by 4.86%, 3.97%, 23.68%, and 18.07%. Also, looking at the power spectrum of the lift coefficient shows that this control approach can stop the vortex shedding and oscillations of the aerodynamic effect quite well. This study shows that the reinforcement learning-based migration control approach can work well in complex unsteady flow fields. It also gives us new ideas and theoretical support for making active flow control of tandem-wing aircraft fast and effective.