A structured model reduction method for large scale networks

Mathematical models of networked systems often take the form of a set of complex large-scale differential equations. Model reduction is a commonly used technique of producing a simplified, yet accurate, description of these systems. Most available model reduction techniques require state transformations, which can cause the structural information of the system to be lost. In this paper, a systematic methodology is proposed for reducing linear network system models without employing state transformations. The proposed method is based on minimising the Hankel error norm between the original system and the reduced order model while ensuring that the state vector in the reduced model is a subset of the original state vector, which preserves the model structure. An error bound between the original and reduced models is ensured and the steady-state behaviour of the system is also preserved. The methodology can be automated so that it be applied to large scale networks. The proposed method can be extended to uncertain systems described by linear parameter varying models. The effectiveness of the proposed methods is demonstrated through simulation examples.