Comparison of linear identification techniques for an open-channel irrigation system

This paper deals with modeling and identification of a real irrigation system (Bourne-France) on the basis of real time-series. The black box and knowledge based modeling with various identification methods, iterative and non-iterative, are tested on real data. Between these methods, the recursive Kalman filter ensures the tracking of the systems parameters (time varying system with time delay), and the subspace methods (non-iterative) are always convergent, numerically stable and the system orders do not have to be known in advance.