Parameter Tuning Methods In Model-free Adaptive Control

ABSTRACT:As a simple and effective advanced control strategy, model free adaptive control does not rely on plant model, and thus has been widely applied in many industrial processes. However, there is no one systematic tuning method for the key parameters of the designed controller, this affects its application to practical systems. Therefore, it’s significant to research on model-free adaptive control parameter tuning methods with systematic tuning principles and algorithm, as same as the tuning method of traditional PID parameter. Furthermore, promoting the proposed method used by the engineering staff has important academic significance and application value.Based on the similarity between model-free adaptive controller and PID controller in structure, this paper references the mature PID parameter tuning method and researches on four model-free adaptive controller tuning methods. The main work is as follows:Firstly, insight into the model-free adaptive control system is researched. Comparing the PID control algorithm and model-free adaptive control scheme based on full form dynamic linearization, it is found that full form model free adaptive controller is similar to incremental PID controller in structure. This provides a theoretical basis to applied PID parameter tuning method in model-free adaptive control system.Secondly, the full form model-free adaptive controller is derived when some pseudo-orders are known. Learning from PID controller tuning methods, four model-free adaptive parameter tuning methods are proposed which include parameter tuning method based on Z-N empirical formula, parameter tuning method based on critical ratio, parameter tuning method based on Cohen-Coon and optimal tuning method based on system performance.Finally, the feasibility of parameter tuning methods is validated through a large number of simulations. Simultaneously, the model-free adaptive controller parameters tuning rule is summarized through the varying parameters of first-order and second-order system model.