To control vibration of support cantilever installed model in real time in wind tunnel, a support cantilever structure integrated with piezoceramics stacks, used as damping elements, is designed. And a neural network proportion-integration-differentiation (PID) real-time control method is proposed. The dynamics equation of the support cantilever is established, and the state equation for system identification is deduced through the neural network. After that, the controller is designed and realized by programming based on Labview software. Finally, the experiments in wind tunnel are implemented to validate the effectiveness of the controlling method. The experimental results indicate that real-time vibration reduction can be executed by the control system. For vibrations excited in wind tunnel at different wind speeds, the root mean square (RMS) amplitudes of vibrations are decreased by more than 55% using the control system in real time. The control method is possessed of robustness, reliability and fault-tolerance.