2015-07-21
1023
System identification has arisen in different areas of application where the system model is completely unspecified but one wants to predict the system response, to regulate the system, or to simulate the system. The only data available are a sequence of known input and a sequence of noise corrupted output. The intermediate objective is to specify a model which agrees with the statistical data. To perform system identification requires three steps: structure determination, parameter identification, and model verification. Before solving the parameter identification problem, one would address the problem of identifiability of parameters. With an assumed structure, is it ever possible to identify the unknown parameters by extracting information from deterministic input and stochastic output data? The capability of answering this question will facilitate the selection of an appropriate model structure. Clearly, one would not select a model structure whose parameters cannot be identified. Thus the question of parameters identifiability is central in the procedures for system identification.
Артикли
