Control Loop Performance Assessment With Subspace Identification

The performance of a controller is crucial to the safety and efficiency of industrial production.In real-world industrial processes,control systems often experience performance degradation over time due to insufficient maintenance.Control performance assessment technology can evaluate and quantify a controller’s performance during operation,determine whether performance is degraded,and offer guidance for controller enhancement and optimization.The rapid advancement of computer processing power has made subspace identification techniques,previously deemed computationally burdensome,a valuable tool for control performance evaluation and optimization.This thesis investigates several control performance assessment methods based on subspace identification and data-driven approaches,with the following main research contents:(1)A method for performance assessment of PID controllers based on subspace nonparametric models is proposed.The matrix of subspace identification intermediate processes is primarily employed,providing the necessary nonparametric model for computing the optimal performance of PID controllers.First,the explicit expression of controller performance concerning PID parameters is deduced using the subspace matrix equation.Next,the subspace matrices corresponding to the process model and random disturbance model are identified using closed-loop data with set point excitation.The estimated dynamic matrix is directly applied in the best performance calculation to obtain the optimal controller parameters.Lastly,the proposed method is demonstrated through numerical simulations and a pilot-scale plant application.(2)Given the complexity of noise model identification and model mismatch in calculating the minimum variance(MV)performance index in multivariable control systems’ performance assessment,a method for estimating the minimum-variance performance index in MIMO feedback control systems without fitting a noise model is developed.First,a multivariate time delay(MTD)matrix is constructed using the process’ s unitary interactor matrix to filter the system output.Second,the minimum variance control law under state feedback is derived using the state space model of a closed-loop system.The performance assessment of a multivariate system is realized by projecting output data,suggesting that the minimum variance output space can be regarded as an optimal subspace of the general closed-loop output space.Lastly,the method’s effectiveness is verified through the numerical simulation case of a multivariate non-square system and a variable delay system.(3)For supervisory control systems,the optimal design of the linear quadratic Gaussian(LQG)controller and performance assessment method based on the LQG benchmark are examined within the framework of the subspace model.Initially,an advanced controller is employed in the supervisory control layer atop the regulator to form a supervised-regulatory control structure,and three potential scenarios of LQG controller design in the cascade control structure are discussed.Then,the control law based on the subspace model is derived for each case,and the LQG performance assessment benchmark calculation method is provided within the subspace framework.Finally,the supervisory LQG controller design and performance assessment based on the LQG benchmark are implemented for the control process of a distillation column.