This dissertation examines the challenge of (i) estimating the internal states of lithium-sulfur (Li-S) batteries based on an experimentally-parameterized multi-physics model, and (ii) optimizing the discharge trajectory to maximize the energy release of a Li-S battery over a fixed time horizon. This research is motivated both by the potential of Li-S batteries to provide higher energy densities compared to traditional lithium-ion batteries and the potential of model-based estimation/control to improve the performance of a Li-S battery. Previous work in the literature optimizes the materials used in Li-S batteries for better performance, and also develops multiple types of models for these batteries, ranging from simple equivalent circuit models (ECMs) to sophisticated diffusion-reaction models. The dissertation builds on the insights from the existing literature, and focuses on the control-oriented study/analysis of Li-S batteries. From the battery management perspective, the essential components include a computationally tractable model, an internal state estimator, and eventually an optimal control strategy. This research addresses the gaps corresponding these key components.

First, this dissertation explores the problem of parameterizing a zero-dimensional physics-based Li-S model. Due to the dependency between the parameters, a simulation-based sensitivity study is performed to provide guidance on the choice of to-be-identified parameters. These parameters are identified by fitting the simulated voltage profile to the experimental data for four models considering different reaction chains. The best fitted model is suggested for the following state estimation study.

Second, there is a need for online state estimation algorithms that take into account the multiplicity of active species in Li-S batteries. This dissertation addresses this gap by developing a model-based unscented Kalman filter for state estimation in Li-S batteries, using the parameterized zero-dimensional model. Simulation-based analysis and study is performed to validate the estimator. This uncovers fundamental insights regarding the observability of Li-S battery states, particularly in the low plateau region.

Third, this dissertation demonstrates the fundamental insight that battery SOC estimation accuracy can benefit from the dependence of battery resistance on SOC. Fisher information is used for developing this fundamental insight, based on a first-order equivalent circuit battery model. Moreover, experimental data from laboratory prototype coin cells are used to parameterize the equivalent circuit model, and the model is utilized in a Monte Carlo simulation study to support this theoretical insight. The simulation study shows a 50% improvement in SOC estimation accuracy in the low plateau region, where the slope of open-circuit battery voltage with respect to SOC is particularly shallow.

Fourth, this dissertation examines the problem of optimizing the discharge trajectory of a Li-S battery to maximize its energy release over a fixed time horizon. This optimization study utilizes a coupled thermal/electrical equivalent circuit model, based on the existing literature, that captures the dependence of battery resistance on both temperature and state of charge. The optimization problem is solved using direct collocation. Simulation results show that trajectory optimization improves total energy delivery over a 2-hour time window compared to both constant-current and constant-power discharge.

The overall outcomes of this dissertation include insights/algorithms that can be implemented into battery management systems to improve the performance of Li-S batteries. These outcomes cover model parameterization/reformulation, state estimation, and optimal control.