Koopman-based Optimal Control of Boost DC-DC Converter

This article presents a global linearization strategy based on Koopman operator theory for a switched boost converter. Applying the Koopman operator approach, any nonlinear process is represented as an infinite-dimensional linear dynamical system. In this contribution, a neural network is used to find a finite-dimensional approximation of Koopman subspace from the measurement data. Then a linear-quadratic regulator (LQR) control framework is applied. The proposed approach is purely data-driven and does not require any prior system modeling. The proposed method was successfully validated in a simulation study of a DC-DC boost converter and compared to conventional LQR control based on operating point linearization.