Self-tuning of PID controller based on process operational data

In this work, a novel algorithm for the self-tuning PID parameters by using process operational data is proposed. A feasible data set is achieved on the basis of dynamic characteristic of PID control loop. With defined ε-insensitive loss function and identification confidence function, the valid data set for model identification is selected from the feasible data set. The valid data set is used to model process objects and then a PID parameters tuning method is given which has a multiple models structure. By using developed optimal stochastic Luus-Jaakola algorithm, an optimized PID controller is obtained which has a perfect effect for all the models. The work of the algorithm does not require staff participation, and the obtained PID parameters can be directly applied to the control loop without a further adjustment. This algorithm has been used in many production plants successfully, and produced satisfactory results. The actual results show that this is a practical algorithm of PID parameter tuning with obvious advantages of simple use and convenient promotion.