Improvement And Application Research Of Process Data Reconciliation Method

Accurate and reliable measurement data can be obtained by data reconciliation, which could provide important decision making basis for computer process control, simulation, optimization, production management. Steady-state detection and gross error detection is research topic in the field of data reconciliation technology. Based on the comprehensive analysis of the traditional methods, some questions about steady-state detection and gross error detection in the field of data reconciliation technology were researched in the dissertation. A new method called the pairwise synchronous detection (PSD) method was proposed to detect and identify gross errors efficiently. A modified method which combined the PSD method with the measurement test method was proposed to reduce the calculation amount. The main contents are as follows:(1) Application research of steady state detection. Based on the principle of the mathematical theory of evidence (MTE) method and the filtering method, the influence on steady state detection of filter factors and the interval length was studied. The ability of steady-state detection and the reliability of practical application of the two methods were compared in the real process. The results show that the filtering method of steady-states detection needs little calculation amount and has better power of test than the MTE method, however, both of them could be used to detect steady-state of the process in chemical industry.(2) The pairwise synchronous detection (PSD) method was proposed. Since the type I errors occured in the traditional method, the GLR-NT test method which combined the GLR test method and the node test method was studied to detect gross errors in the measurement on the basis of comprehensively analyzing the present method of gross error detection. Calculation results indicate that the GLR-NT method could decrease the probability of type I error, however, it could not improve the gross error detection rate. In order to improve the gross error detection rate, through recomposing the test statistics, a modified method called the pairwise synchronous detection (PSD) method was proposed to identify gross errors with simultaneous compensation. A simulation example was presented to compare the performance of the modified method with the GLR test method on the same level of significance. The practical application shows that the PSD method has higher value of overall power and expected fraction of correct identification, and lower value of average number of typeⅠerror than the GLR method. Therefore, the new method could improve the gross error detection rate and decrease the probability of type I error effectively. What's more, it could give more accurate compensation value for the measurements.(3) The PSD method was applied to nonlinear system. By means of the linearization technique of constraint equation, the PSD method was applied to the nonlinear system. Jacobian matrix and the test statistic was computed through the constraint residual of the pair-wise variable. Then the gross error in variable was detected if the statistic value higher than the critical value. The case study proved that the PSD method could be used successfully than the GLR method in the nonlinear process.(4) The MT-PSD strategy was proposed. The MT-PSD strategy was prensented which combined the PSD method and the measurement test method to reduce combinational numbers of variables and calculation amount of statistical tests. The new strategy was applied to practice of gross error detection and data reconciliation in the industrial process. The application study shows that the performance of the MT-PSD strategy has good enough for gross errors detection, so that it could be applied effectively to the industrial process.