Identification Of Steady-State Models For Bilinear Industrial Processes

With the fast development of distributed computer control systems for on-line computer control of continuous production processes , determining and maintaining the optimum steady-state operating conditions on-line of the industrial process is playing the more and more important role in chemical, metallurgical, petroleum and power industries etc. As is well known, steady-state models play a very important role in steady-state optimizing control. Researchers have proposed some steady-state identification techniques. For bilinear industrial processes and bilinear large-scale industrial processes, this thesis improves the identification techniques which now is available and proposes some new identification techniques, by which the steady-state models are obtained under very mild conditions. Simulation studies are provided, and the result shows that the presented techniques are very efficient.For steady-state behavior of bilinear industrial processes which are described by static models, a new identification technique is presented which uses the step signal of the step-points changes as the input identification signals in the course of optimizing control, and the steady-state models are obtained. The new technique converts the identifying of unknown matrixes of steady-state models into solving matrixes algebraic equations.For bilinear industrial processes which are described by dynamic models, a new identification technique is presented, by which the steady-state models are obtained and the estimates are strong consistency. In the thesis , the identifiability for steady-state models is studied and gives the sufficient conditions for system identifiability.This thesis proposed a decentralized identification technique, by which we can obtain the strong consistency estimate for the bilinear large-scales industrial process. The novel technique can not only use step signals of set-points as input identification signals, but also the identifications of the input-output models of sub-processes are implemented in the" corresponding local units, hence the disturbance of the processes and the interchange of information of sub-processesare largely decreased. So the technique is simple and has high identificationprecision.