A new system identification scheme using modified Orthogonal Forward Regression and errors-in-variables techniques

Abstract

System identification is one of the most important parts in science and technology. This branch of science specially has an important role in control engineering, because with a proper identification method one can model a phenomenon to control its performance. In recent years, there has been a rapid development of process control techniques. However, there is currently no ultimate solution to the system identification problem. In order to solve the problem mentioned above, a new system identification scheme for time varying systems is developed in this thesis to improve the performance of current identification algorithms. Also, two errors-in-variables system identification techniques have been discussed and compared. As compare to current OFR algorithms, the new system identification scheme has made the following improvements. The first improvement is that the new system identification scheme can now detect the system parameters correctly when input and output data are both corrupted by noise, while the classical OFR algorithm cannot achieve this. The second improvement is that the new system identification scheme requires less information than the modified OFR algorithm.