Immersion Based Nonlinear Observer Design for Power Systems

Abstract

Geometric techniques for observer design operate by transforming the original system, through a change of coordinates into a form allowing an efficient observer design. This change is predetermined up to a certain extent by what outputs are measured and therefore challenging to find. Lately, Koopman-based observers have gained popularity. These represent a more generalized approach of embedding the original system into a higher dimensional linear system. A major challenge in using this framework is the choice of new higher dimensional coordinates allowing for a possible linearization. Luckily, high gain observers make it possible to relax the linearity requirements. This work explores the use of multiple, decoupled, high gain observers to estimate the internal states. The results are demonstrated on the observer design of a classical machine.