Data Coordination And Gross Error Detection Method And Its Application

Process measurement data is the basement of process control, optimization, operations analysis as well as management. Therefore, accurate and reliable measurement datas is the cornerstone of modern industrial processes. However, in the actual measurement process, because of various uncertain sources such as measurement error, instrument failure and equipment leaks and so on, the actual measurement datas contain random errors and/or gross errors inevitably, what affect accuracy and reliability of the datas seriously.By using the redundancy in process measurement datas,the purpose of data reconciliation and gross error detection is minimizing the effects of errors and make the measurement datas obey the relation of the internal energy balance, mass balance, and other physical and chemical conservation laws to get the adjusted values, which is reasonable and close to the true value, and try to estimate the unmeasured variables.So far,the researches on data reconciliation and gross error detection include conventional data reconciliation and gross error detection, simultaneous data reconciliation and gross error detection.Among them, the essence of the second aspect research is to introduce the robust estimation theory. This dissertation studies the principles and methods of data reconciliation and gross error detection systematically, and the main contents are as follows:(1) Research the NT-GLR(Nodal Test-Generalized Likelihood Ratio) in-depth, which is a way of data reconciliation and gross error detection. The method is achieved by NT detection and GLR location and compensation. For the problem of inaccurate compensation values, an improved method of NT-GLR is proposed. After detecting all the gross error streams, which are acted unmeasured variables and estimated by Least Squares estimation data reconciliation algorithm, and more accurate reconciliation effect is acquired.(2) Research the robust estimation principle, a new robust estimation function is proposed, and a new simultaneous method of data reconciliation and gross error detection is proposed, which is insensitive to the abnormal data which from the normal data,and can achieve data reconciliation and gross error detection simultaneously and obtain accurate and reliable reconciled values.This robust estimation method is used for steady-state system and dynamic system,the simulation results verify the effectiveness of the new robust estimation method.(3) Research the particle filter algorithm, a simultaneous method of data reconciliation and gross error detection is proposed, which bases on particle filter and robust estimation.For the method of the dynamic data reconciliation and gross error detection based on particle filters, which has the degeneracy problem of particles, the robust estimation is introduced into the particle filter, using the robust function to update the particle’s weight secondly,then particle filter secondly, in this way, particle’s confidence is strengthen and the degeneracy problem of particles in the dynamic process based on particle filter can be avoided.The efficiency of the proposed approach is proved by simulation results,which is studied on dynamic linear system and dynamic nonlinear system.(4) The methanol production process of Shanghai Coking Company is researched, and a simplified material static model is builded.By using the robust estimation algorithm(WRE) proposed in the paper and nonlinear programming(NLP) method to achieve dynamic data reconciliation and gross error detection for the process measurement datas, and obtain the coordination results of total flow and its composition and estimates based on the theory and ideoligical of two-step method.