The invention discloses a self-learning-based rotation control method for a dangerous condition of a tower crane. The self-learning-based rotation control method particularly comprises the following steps of: 1, carrying out self-learning on the inertia and a rotation running characteristic of the tower crane on the tower crane; 2, carrying out deceleration or shutoff control as appropriate according to the current running state of the tower crane; 3, after low-speed cut-off control of the tower crane, executing closed-loop control, if the dangerous angle of the tower crane at the time is more than an ideal parking position, re-starting low speed in a control direction of the tower crane by a system, allowing the tower crane to continuously advance at a low speed to be close to a target position; and 4, immediately braking when the tower crane enters a red line due to out of control, directly stopping the tower crane, delaying a brake for a certain time, then opening the brake and allowing the tower crane to rotate towards a safety direction. According to the self-learning-based rotation control method, the rotation of the tower crane is passively controlled, the parking regulation and braking control are carried out to ensure the parking stability, and the impaction of the rotation inertia of the tower crane is avoided.