Reduction of Power System Dynamic Models Using Sparse Representations

Abstract

This paper proposes a model reduction technique that simplifies the dynamic equations of complex power networks, using sparse representations of the system matrices. Instead of removing components from the state vector, elements from the system matrices are eliminated such that these matrices become sparse. This is achieved by three different numeric algorithms that approximate the original system model using fewer nonzero elements. These algorithms lead to simpler models, since the complexity of operations involving sparse matrices is primarily affected by the matrices density. Furthermore, this approach enables to identify significant dynamic relations between units in the network. The proposed methods are demonstrated on several test-case systems with 9 and 2383-buses. In these examples, more than ${\text{90}}$ % of the elements in the system matrices are eliminated.