Data Reconciliation And Steady-State Identification Of Chemical Process

Chemical products are playing a more and more important role in people’s daily life as with the continuous development of society. At the same time, Our country is gradually becoming the "World Factory", as a result, The chemical industry will certainly become more and more significant among all the industries. That means, it is a very meaningful subject to research how to implement optimization on chemical process to improve production without increasing energy consumption and pollution of the environment.High accuracy, high reliable data collected from steady-state has a decisive impact on the implementation of optimization, if the optimization is implemented based on data that are not accurate or reliable enough, Not only the optimization will turn out to be a failure, but also an irreversible damage may be done to the whole chemical process. Data reconciliation and steady-state identification are technologies that are capable to provide process optimization with high accuracy, high reliable steady-state data.The first part of this paper discussed the basic theory of data reconciliation, and an aborative research focused on the key and difficult part in data reconciliation, gross error processing, is fully implemented. Data reconciliation based on robust estimation, due to its ability in managing the two types of errors synchronously, is also integrally reviewed. And two kinds of robust objective functions are proposed according to the robust estimation theory. The efficiency and effectiveness of both robust objective functions in data reconciliation are detailedly illustrated by two examples.Two commonly used steady-state detection technology based on filtering method are reviewed and studied, two defects are found in the learning process of the classical F-test, then an improved F-test is put forward on the basis of this two defects,three standards are proposed as to evaluate the efficiency of different steady-state detection methods, at last the efficiency of the improved F-test in steady-state detection is confirmed by two examples.Because of its massive advantages, CSTR is widely used in chemical industry. This paper also adopts a CSTR as the object model in examining the research achievements above. By combining the technology of data reconciliation and steady-state detection, high accuracy, high reliability steady-state data can be extracted. This lay a solid foundation for system modeling, process optimization, parameter estimation and other follow-up works.