Improved PID Control Based On RBF Neural Network

It is difficult to get classic PID controller precise parameters, and PID control is based on the precise mathematical model. So PID control is not adaptive enough in nonlinear and time-variant control systems. The neural network have a strong self-learning ability to adapt to complex environments and multi-objective control requirements, and can approach any nonlinear continuous function with any precision. The purpose of this paper is to find a new method to merge the advantages of both.First, a new control algorithm of RBF identification network is proposed in the paper. According to RBF network identifying the discrete model of plant on-line, the Jacobian information can be got. Then , adjusting PID parameters adaptively on-line with BP network. This method overcomes the adverse effect generated by the uncertainty, and solves the problem that the traditional PID control method has poor robustness and is limited to the accurate mathematical model .A second-order nonlinear system is simulated in the paper ,the result shows that the control algorithm has better adaptability and robustness, and has the advantages of higher anti-interference ability and adaptability to parameters'changing than conventional PID control.Second, this paper briefly describes the composition of magnetic suspension system and working principle and analyses the system dynamic model in detail, which notes that the system is unstable and non-linear. In this paper, a nonlinear dynamic model of maglev is established using S-functions in MATLAB environment. Using self-learning, adaptive features of RBF neural networks, a adaptive PID controller is designed based on the traditional PID controller.At last, an improved adaptive PID controller is built based on the traditional PID controller, which solves the dynamic and static performance requirements of magnetic Levitation Ball control system. The results showed that the improved adaptive PID controller based on RBF on-line identification has better adaptability and robustness compared with conventional PID controller.