Modelling, state estimation & optimal charging control for a lithium-ion battery

© 2016 Dr. Changfu ZouLithium-ion (Li-ion) batteries are well-suited chemistry technology for a myriad of applications such as portable electronic devices, electrified vehicles and energy storage in power systems. To satisfy consumers' demanding requirements, the batteries are desirable to have high energy/power density, long life cycles and fast charge capability but provide these at affordable costs with safe and reliable operations. However, the current batteries are limited in one or more of these characteristics. This thesis leverages physics-based models and model-based estimation/control techniques to improve battery performances. The first phase of this project focuses on battery model simplification. A physics-based Li-ion battery model capturing the coupled electrochemical, thermal, electrical and ageing dynamics is first developed. Through identification of disperse time-scales inherent in the initial model, a framework for PDE battery model simplification is proposed by using singular perturbations and averaging theory. The novelty arises from the systematic procedure for model simplification through which the underpinning assumptions can be explicitly stated. As a consequence, libraries of simplified battery models with interconnections are obtained and applicable for a range of applications. Simulations are conducted to validate the low order models against the initial model. The capability of the proposed modelling framework is demonstrated on a model-based evaluation procedure for charging strategies. With appropriate battery models obtained, model-based algorithms are developed to estimate the battery state-of-charge (SOC) and state-of-health (SOH). First, a nonlinear estimator synthesised from an extended Kalman filter (EKF) and a reduced-order battery model from the proposed model simplification framework is designed to simultaneously estimate the SOC and SOH. Second, a novel multi-time-scale estimation algorithm is proposed for a class of singularly perturbed systems and is pertinently applied to estimate battery SOC and SOH in different time-scales. This reduces computational burden and avoids potential issues such as ill-conditioned gains and convergence. Illustrative results show that the multi-time-scale observer is effective in estimating both the SOC and SOH despite a range of common errors due to modelling, initialisation and measurements. Finally, by utilising a computationally efficient model from the above model simplification framework and the developed multi-time-scale estimator, a novel model predictive control (MPC) based charging algorithm is proposed for a Li-ion battery to balance the competing objectives, i.e. battery's health and charging time. Specifically, the charging strategy is formulated as a linear-time-varying (LTV) MPC problem for tracking SOC and SOH reference signals, in which constraints on the input current and battery internal states are explicitly considered. The capability of the proposed charging strategy is demonstrated via high fidelity simulations. As compared to its industry benchmark, i.e. the CCCV approaches, the MPC-based charging is found to have the ability to provide longer battery lifetime yet requiring less charging time. Furthermore, the proposed algorithm offers battery users the freedom to specify battery lifetime or charging time based on their requirements. The developed modelling, estimation and control algorithms are very general so as to be applicable to different types of Li-ion batteries and cell chemistries