Dynamic–quadratic balancing: A computational approach to balancing and model reduction for affine nonlinear systems

Abstract

In this paper, we discuss a new approach to balancing (known as dynamic–quadratic balancing) and model reduction for affine nonlinear system. We give a fresh look to balancing in terms of the dynamics of the system, rather than simply a structural property. With this perspective, we also develop a new approach to obtaining the balancing transformation in one step, instead of a three-step process as proposed in earlier methods. Further, we explore the relationship between quadratic balancing and input–output stability. In addition, we also develop a computational approach for obtaining the balancing transformation by solving a coupled system of PDEs or inequalities (Lyapunov/Hamilton–Jacobi type). After that, model reduction can be carried out in the conventional way using Hankel singular-value functions or using a new criterion. Finally, we present some examples to clarify the results.