Performance Evaluation And Optimal Tuning Of Pi Controller For Typical Thermal Process In The Case Of Model Plant Mismatch

PI control is widely used in thermal process control.After the control loop is put into operation,the characteristics of the controlled object will change with the running time,which makes the process model describing the characteristics of the controlled object(referred to as the actual model)no longer match the process model used in the original controller design stage(referred to as the design model),which is called model plant mismatch.Model plant mismatch may cause adverse effects on the control performance,which may lead to the degradation of control quality and threaten the safe operation of thermal equipment.Therefore,it is necessary to identify the process model again,evaluate the influence of the control performance quantitatively and adjust the controller to improve the control performance after the model mismatch.In this paper,the optimization research is carried out for the above three aspects.The main contents and achievements are as follows:(1)Aiming at the problem of process model identification in the case of model plant mismatch,a model identification method based on BP neural network regression modeling is proposed.This method defines the characteristic variables of the closed-loop set point step response,and establishes the relationship between the process model and the above characteristic variables by using BP neural network.According to the established relationship,the mismatched process model can be identified by using the closed-loop response eigenvalues.Simulation results show that the proposed method has better identification accuracy and stability than the traditional model identification method based on BFGS or GA optimization algorithm.(2)Aiming at the problem of performance evaluation of PI controller in the case of model plant mismatch,a quantitative evaluation method for the influence of model parameter change on control performance is proposed.In this method,the relationship between the model mismatch parameters and the control performance index is deduced to obtain the control performance index after the model mismatch,and then the influence of the model mismatch on the control performance is evaluated quantitatively by the way of performance index normalization.The evaluation results can be used as the basis to judge the mismatch severity of the model.Compared with the existing research,the proposed method extends the applicable object from the first-order plus time delay model to the second-order plus time delay model and the critically damped second-order plus time delay model,and increases the evaluation range of control performance to three aspects: set point tracking performance,load disturbance rejection performance and robustness.(3)Aiming at the problem of PI controller tuning in the case of model plant mismatch,a tuning method of PI controller aiming at the compromise optimization of control performance is proposed.The controller tuning method is transformed into the optimization problem based on genetic algorithm,and the objective of optimization is to make the two control performance indexes approach the optimal value that they can achieve as much as possible,and the robustness is taken as the constraint condition.Compared with the existing research,the proposed method can make the performance of set point tracking and load disturbance rejection achieve better compromise effect,and the control performance is improved to a greater extent than before.