For realizing the autonomous dynamic soaring maneuver without relying on known wind shear profiles, this paper develops a real-time trajectory correction approach. The strategy is inspired by the fact that large seabirds do not know the wind profile and yet they presumably can feel the wind variations and predict their own motions. Utilizing only the instant wind information rather than the complete profile, this approach formulates a strictly convex quadratic programming problem based on the instant wind strength increment and the predicted errors to determine bounded control inputs. Initialized by an energy-neutral dynamic soaring trajectory that requires the minimum wind strength, control inputs can be corrected in real time. Simulation results demonstrate the effectiveness of the proposed approach under different unspecified wind profiles.