Simultaneous improvement of control and estimation for battery management systems
Abstract
The state of charge of battery systems is an important metric typically estimated by observation models, represented by open-circuit voltage graphs. These observation models are often nonlinear in the state of charge, resulting in varying observability from a state estimation perspective. In this paper, we employ a stochastic optimal control (also known as dual control) approach to simultaneously satisfy the control objective in the state of charge of battery systems and improve estimation accuracy. This is achieved implicitly by prioritizing trajectories that pass through high-observability regions of the state space, thereby improving the quality of future measurements. We apply our algorithm to a numerical simulation of a multi-battery system and show a statistical improvement in both the control objective and the state estimation error.