Research On Data Reconciliation And Its Application

Reliable process data is the foundation of efficient process control, process operation and evaluation of process performance. However, Process measurements inevitably contain random errors and sometimes even contain gross errors. By using the redundancy in process measurements, data reconciliation can be used to eliminate measurements with gross errors, reduce the effect of random errors and make measurements comply with the conservation laws, such as the conservation laws of mass and energy balances. And therefore, the unmeasured data can be estimated. This dissertation studies the problem of data reconciliation systematically, and makes progresses in the following aspects:1. Iterative Measurement Test (IMT) method is an effective gross error detection method. But least square method is used in IMT to attain reconciled results to construct statistical value and therefore gross errors can be easily misidentified. Under the condition that measurement errors of process variable are correlated, a modified IMT method is introduced which can increase the power of correctly identifying gross error.2. A modified Serial Identification with Collective Compensation (SICC) method is introduced. By using temporal redundancy, upper and lower bounds of process measurements are added in the modified SICC method to avoid the misidentification of gross errors. To prevent the computation of project matrix, measurements are compensated by using the estimated gross error magnitudes after each gross error is identified. And after that, Measurement Test (MT) is used to find the gross error candidates. Necessary cycle detection is added to avoid the singular matrix appearing after gross error compensation,. Simulation results verify the effectiveness of modified algorithm.3. Based on the analysis of the cycles for sensor networks, mathematical representation of redundancy degree is introduced. To avoid infeasible solution, bound of redundancy degree is analyzed to give theoretical guides for the design of sensor networks. Based on a combination of concepts drawn from graph theory and Integer Linear Programming (ILP) methods, nonredundant sensor networks with minimum cost and redundant sensor networks that satisfy constraints related to redundancy degree of key variables, are established. The obtained method is verified by simulation results.4. A new method to solve robust data reconciliation problem of nonlinear chemical process is proposed. This method is very convenient in computation. Byusing several technologies including linearization method, penalty function, virtual observation equation and equivalent weights method, the robust data reconciliation problem can be transformed into a least squares estimator problem. Simulation results for a nonlinear chemical process demonstrate the efficiency of the proposed approach.5. Based on the analysis of the limitations for the existing methods, a modified approach of dynamic data reconciliation and outlier detection is presented. This method can use more information of normal data, and can efficiently decrease the effect of outliers. The simulation results on a CSTR process verify the effectiveness of the obtained algorithm.6. The application of Dempster-Shafer theory in gross error detection is discussed. The situation when leak appears in the process is considered. The environmental node is introduced as one proof in the Dempster-Shafer theory. The result is verified by simulation.7. For the problem of mass flow data reconciliation in Anqing petrochemical company, the procedure of building an efficient and user-required linear account balance model is discussed in detail.The dissertation is concluded with a summary and prospect of future data reconciliation researches.