Studies On Gross Error Detection Based On Data Reconciliation

As the characteristic information of plant operation state, chemical process measured data is the foundation of CIPS (Computer Integrated Process System). Data reconciliation technology plays a very important role in the operation and control of chemical plant. As the most important component of CIPS, process data reconciliation technology is one of the main means to acquire high quality data. Many works have been done in this field and much progress has been made in the last few years. The research on the data reconciliation technology and its application has important theoretical and practical significance in the optimal control and management of process plant.On the basis of analyzing the present method of data reconciliation, a strategy which combined the data reconciliation method with the gross error detection is studied. The result of the data reconciliation with the inequality constraints was used to calculate the statistical value to detect the gross error. A method which combined the data reconciliation using inequality constraints with the traditional MT-NT method was presented. In order to detect and identify more than two gross errors efficiently, a strategy which combined the ANN (Artificial Neural Networks) with the statistical test method was proposed.(1) For the cases of process model with uncertainties and equipment leakage, a data reconciliation method with inequality constraints and its application examples were presented. Compared with the traditional method with equality constraint, the result shows that this method was of more extensive application. Even in the case of leakage, an accurate result could still be obtained by inequality data reconciliation method, and in the case of process model with uncertainties, a more rational reconciliation result could be acquired.(2) On the basis of the performance of data reconciliation with the inequality constraints, the gross error detection method was presented which makes use of the results to set up the statistical value and then detects and identifies the gross error. The practical application results shows that this method can weaken the spread of the gross...