Research On Data Reconciliation And Application

By using the redundancies of process measurements and statistical methods, Data rectification technique can eliminate gross errors in measurements, reduce the effects of random errors and estimate the values of unmeasured variables. Finally it can enhance the reliability, the consistency and the integrality of the process datas. In this thesis, decomposition-coordination algorithms of large scale systems, gross error detection and the problems in the application of data reconciliation are studied. The main contents of this thesis are described as follows:1. The development of data rectification technique is reviewed from the aspects of data reconciliation, gross error detection and sensor network redundancy analysis. And the applications of data rectification in process industry are introduced.2. The necessity of investigation in decomposition-coordination of large scale systems is discussed and some methods in existence are introduced. A new decomposition-coordination method is proposed. Compared with conventional approaches, this new method doesn't need to use reconciliation parameters, and the iterative calculation procedure isn't needed. So this new method has high efficiency and satisfies chemical industry's requirements.3. Some methods and strategies of gross error detection are introduced. An improved NT-MT method for gross error detection is proposed. The improved NT-MT method can decrease iterative times and avoiding from misjudging the gross errors effectively.4. Some problems in the application of data reconciliation, including the estimation of measurement error variances, nonlinear data reconciliation are discussed. And some results are given. The approach for gross error detection based on neural network is introduced. Compared with conventional approaches which are based on optimization technique and statistical methods, This method has the advantages of simpleness, fitting online use and dealing with nonlinear problemeffectively.Finally, a summary of this thesis is given, and the perspective of data reconciliation is prospected.