Performance Assessment Of Control Systems Based On Minimum Variance Benchmark

Control Performance Assessment(CPA)is an effective tool in the industrial process that can be used to estimate the performance of feedback control loops.Since the CPA technique based on the minimum variance benchmark is only applicable to the control system driven by the stationary signal,the performance evaluation method can not be applied directly when the disturbance is non-stationary.However,in practical industrial production,the perturbation ends are mostly non-stationary signals,such as sinusoidal signals and white noise or step signals and white noise.On the univariate system,there is a non-stationary signal,which results in the closed-loop output instability of the system.Therefore,a performance evaluation algorithm based on the minimum variance benchmark is proposed.The closed-loop output data routinely run by the control system can accurately evaluate the performance of such systems.The method is as follows:The closed-loop output of the control system is stabilized by differential method,and then the stabilized output are modeled and analyzed.The impulse response coefficient of the closed-loop transfer function is calculated by the correlation analysis(FCOR)for differential control system.The impulse response coefficient of the control system closed-loop transfer function only under white noise can be reduced,and the performance index of the system is calculated.At present,the performance evaluation of the multivariable system based on the minimum variance criterion often needs to know several markov coefficient matrix of process transfer function or interactor matrix.And this generally requires the identification of multivariable systems.However,the identification of multivariable systems generally introduces additional continuous excitation signals at the set value of the closed-loop system,which can result in destroying the normal operation of the system and deteriorating the system operating performance.For the calculation of several markov coefficient matrix of process transfer function or interactor matrix,it is necessary to introduce the excitation signal on the set value of the closed-loop system to identify the system.In this paper,an on-line performance evaluation method based on the minimum variance criterion is proposed without additional Continuous excitation signal that can be applied to the control system to identify the system and the calculation of the interactor matrix to solve the multivariable system performance indicators.The method is as follows:Firstly,the system output sequence is obtained by changing the parameters of the multivariable system controller.Secondly,according to the above sequence,markov coefficient matrix of the closed-loop transfer function can be calculated.Finally,the system performance index can be calculated.