Research on Neural Network PID Quadratic Optimal Controller in Active Magnetic Levitation

ABSTRACT

In response to the uncertainty, nonlinearity and open-loop instability of active magnetic levitation control system, a neural network PID quadratic optimal controller has been designed using optimum control theory. By introducing supervised Hebb learning rule, constraint control for positioning errors and control increment weighting are realized by adjusting weighting coefficients, using weighed sum-squares of the control increment and the deviation between actual position and equilibrium position of the rotor in active magnetic levitation system as objective function. The simulation results show that neural network PID quadratic optimal controller can maintain the stable levitation of rotor by effectively improving static and dynamic performances of the system, so as to maintain the stable levitation of rotor in active magnetic levitation system which has stronger anti-jamming capacity and robustness.

Keywords:

Active magnetic levitation, Adaptive PID controller, Quadratic optimal, Robustness, Single neuron.