PID controller design of based on neural network and virtual reference feedback tuning

The paper presents a method of data-driven parameter setting based on neural network, which is aimed at the nonlinear controlled objects, and these objects are difficult to establish accurate mathematical model. The network connection weights and node threshold are adjusted to identify the controller parameters by comparison of the virtual reference feedback tuning performance, and this idea can skip the controlled object modeling process. Also, the relationship between VRFT and IMC is derived. In addition, the paper made the proof of neural network learning rate can guarantee the convergence of precise tracking error within limits, and also combined the VRFT parameters to prove the stability of the closed-loop system. Simulation shows that this method has some characteristics, such as strong tracking performance, fast response, good control results for nonlinear plant and so on.