On Model Simplification in System Identification

Abstract: This report deals with the connection between system identification and model reduction. We propose an identification algorithm that is based on the least squares identification method and either of the model reduction techniques: Frequency weighted truncated balanced realization or frequency weighted optimal Hankel-norm model reduction. The advantage of this algorithm's that it has a closed form solution. An overview over these the model reduction methods is given. A more general discussion of the connections between norms in system identification and in model reduction is also given.