Modelling, Estimation, and Cell-to-Cell Heterogeneities analysis in Lithium-Ion Battery Systems

The development of reliable, long-lasting, and safe energy storage systems is crucial for effectively supporting the transition from fossil fuels to renewable energy resources, thereby driving a sustainable global electrification process. However, ethical concerns regarding the materials used in the battery supply chain, as well as the safety and reliability of large battery modules/packs must be properly addressed. One effective commercially mature solution to mitigate the former challenge is the use of lithium iron phosphate (LFP) batteries. Unlike traditional batteries that rely on cobalt, manganese, and nickel, LFP batteries overcome the scarcity and ethical controversies associated with traditional battery materials, offering a more ethical and sustainable alternative. Despite their advantages, LFP batteries present certain technical challenges. The plateaus in open circuit voltage, hysteresis, and path-dependent dynamics exhibited by LFP batteries make them challenging to model, estimate, and control. In this work, the development of control-oriented reduced-order models based on physics-based electrochemical model for LFP batteries is presented, providing a foundation for the development of advanced LFP-based Battery Management System (BMS) strategies. Additionally, a properly designed BMS can ensure the safe and optimal operation of battery modules and packs regardless the cell chemistry, addressing the second challenge. In particular, it is crucial to recognize and accurately quantify cell-to-cell (CtC) variations, as these can significantly impact the overall performance, heterogeneity, and degradation of the battery system. The ultimate goal is to enhance BMS algorithms to account for CtC variations, increasing, in turn, their reliability and robustness. To address this, in this study the effects of CtC variations on parallel-connected battery modules are extensively analyzed using comprehensive statistical methods. Based on full factorial design of experiments, the impact of different sources of CtC variation, such as uneven interconnection resistance, operating temperature, cell chemistry, and aging conditions, are evaluated on the module performance. The collected experimental data form the foundation for developing data-driven models, including interpretable multilinear regression and black-box machine learning models, which are then used to explain and quantify the contribution of each CtC source to the corresponding module-level responses. Further, the analysis is enhanced using a high-fidelity, experimentally validated electrochemical model that incorporates detailed module-level features, considering both short- and long-term implications on module responses. Additionally, leveraging the battery system digital twin, a straightforward cell arrangement strategy to mitigate thermal gradients in parallel-connected battery modules is proposed. This strategy aims to reduce aging disparities between cells, thereby extending the overall lifespan of the battery module. Finally, a state estimation algorithm for individual cells within parallel-connected battery configurations is formulated and validated against experimental data. Utilizing a Moving Horizon Estimation approach, this algorithm provides accurate state estimates, representing a key step toward enhancing battery pack cell balancing and the development of novel fault detection and isolation strategies.