Efficient Lithium-Ion Battery Model Predictive Control Using Differential Flatness-Based Pseudospectral Methods

Abstract

This paper proposes an efficient nonlinear model predictive control (NMPC) framework to solve nonconvex lithium-ion battery trajectory optimization problems for battery management systems (BMS). It is challenging to solve these problems online due to complexity and nonconvexity. To address these challenges, we combine four established techniques from the control literature. First, we represent the single particle model (SPM) using orthogonal projection techniques. Second, we exploit the differential flatness of Fick’s second law of diffusion to capture all of the dynamics in one electrode using a single scalar trajectory of a “flat output” variable. Third, we optimize the above flat output trajectories using pseudospectral methods. Fourth, we employ the NMPC strategy to solve the battery trajectory optimization problem online. The proposed NMPC framework is demonstrated by solving 2 optimal charging problems accounting for physics-based side reaction constraints and is shown to be twice as computationally efficient as pseudospectral online optimization alone.