Nonlinear Model Identification for Thermal Control in BEV: A Data-Driven Approach Using Sparse Identification of Nonlinear Dynamics

Abstract

This thesis investigates the use of data-driven system identification method to support control development for the thermal management system of a battery electric vehicle. The identification process is carried out using the Sparse Identification of Nonlinear Dynamics (SINDy) method combined with sequential thresholding as an optimizer. The goal is to obtain a control model suitable to use for the development of a nonlinear model predictive controller(NMPC). Several models of different complexity and accuracy are identified from recorded data and evaluated offline. To assess their ability to reach a set-point, each model is tested in a single-run optimal control problem using a direct multiple shooting approach.