锂离子电池（LIB）是一种锂离子在正负极之间交换的电化学储能装置。LIB的正极是该电池中最昂贵的组件，占该电池生产总成本的50%以上。与镍和钴基阴极相比，磷酸铁（LiFePO4, 也称为LFP）的阴极具有优越的热和化学稳定性，在高温下不会发生分解，是一种更安全的阴极材料。

钴和镍的缺失为电池供应链的可持续性提供了一条途径，并有助于创建一个道德能源市场。LFP电池通常使用LiFePO4和石墨作为正负电极活性材料。正极的两相变区域会共存富锂相LiβFePO4和一个贫锂相LiαFePO4，进而导致一个平坦的开路电压曲线（OCV，正负电极的开路电位之间的差值）。

Fig. 2 Battery modeling and phase transitions.

平坦的OCV曲线由于缺乏从电压输出测量中对系统的可观测性，从而使估计电荷状态（SOC）的任务具有挑战性。此外，两相变还伴随着明显的滞后和路径行为依赖行为，即对于相同的SOC，电池会根据充电或放电而松弛到不同的OCV值，这给电池管理系统（BMS）策略的设计带来了额外的挑战。

Fig. 3 Hybrid model architecture.

为了设计有效且高性能的BMS，来自美国斯坦福大学能源科学与工程系的Simona Onori教授团队，开发了一种混合模型，将基于物理的建模优势与机器学习模型描述未知物理的能力相结合，以捕捉瞬态操作中的滞后和路径依赖行为。

Fig. 4 Hybrid model performance.

在该混合模型中，基于物理的模型是通过平均核-壳的增强单粒子模型来描述石墨阳极电池中磷酸铁正极的两相变操作，机器学习部分基于对电池在不同充放电模式下的电化学内部状态的了解，从模型特征和实验中学习滞后行为，以补偿模型的不确定性，并分别对19和15小时的电动汽车的真实驾驶轮廓进行训练和验证。

Fig. 5 Hybrid model energetic analysis.

作者的提出的混合模型以数据为中心来开发基于机器学习的伪滞后模型，减少了实验时间，创建了高信息行和低维的数据集，对电池性能分析、合成数据生成以及开发用于BMS设计的降阶模型具有重要意义。该文近期发布于npj Computational Materials 10:14 (2024).

Fig. 6 Training and testing datasets.

Editorial Summary

Battery management system: Physics-based model with machine learning

Lithium-ion batteries (LIBs) are electrochemical energy storage devices where lithium ions exchange between the positive and negative electrodes. The positive electrode of a LIB is the most expensive component of the cell, accounting for more than 50% of the total cell production cost. Compared to nickel- and cobalt-based cathodes, the lithium iron phosphate (LiFePO4, also referred to as LFP) cathodes offer superior thermal and chemical stability, resulting in a safer cathode material that does not decompose at high temperatures. The absence of cobalt and nickel suggests a pathway for a resilient battery supply chain and contributes to the creation of an ethical energy market. LFP batteries typically use LiFePO4and graphite as positive and negative electrode active materials, respectively. The two-phase transition region of the positive electrode coexists with a lithium-rich phase (LiβFePO4) and a lithium-poor phase (LiαFePO4), leading to a flat open-circuit voltage (OCV) curve, defined as the difference between the open circuit potentials of the positive and negative electrodes. The flat OCV curve makes the task of estimating the state of charge (SOC) challenging as it causes a lack of observability of the system’s states from the voltage output measurements. Moreover, the two-phase transition is accompanied by pronounced hysteresis and path dependence behavior, meaning that for the same SOC, the battery may relax to different OCV values depending on whether it is being charging or discharging, posing additional challenges for the design of battery management system (BMS) strategies.

To design an effective and high-performance BMS, a team led by Prof. Simona Onori from Energy Science and Engineering, Stanford University, developed a hybrid model, combining the advantages of physics-based modeling with the ability of machine learning to describe unknown physics, to capture the hysteresis and path-dependent behavior during transient operation. In this hybrid model, the physics-based model describes the two-phase transition operation of the LFP positive electrode through an averaged core-shell enhanced single particle model, while the machine learning component is based on the understanding of the electrochemical internal states of the battery during different charge and discharge operation over several driving profits. It learns the hysteresis behavior from simulated features and experiments to compensate for model uncertainty, and is trained and validated on real-world driving profiles of 19 and 15 hours for electric vehicles, respectively. The hybrid model shown in this study presents a machine-learning-based pseudo-hysteresis model, reducing experimental time, creating high-information-density and low-dimensional datasets. It is of great significance for battery performance analysis, synthetic data generation, and the development of reduced-order models for BMS design.

This article was recently published in npj Computational Materials 10: 14 (2024).

原文Abstract及其翻译

Accelerating the transition to cobalt-free batteries: a hybrid model for LiFePO4/graphite chemistry (加速向无钴电池过渡：LiFePO4/石墨化学中的混合模型)

Gabriele Pozzato,Xueyan Li,Donghoon Lee,Johan Ko&Simona Onori

AbstractThe increased adoption of lithium-iron-phosphate batteries, in response to the need to reduce the battery manufacturing process’s dependence on scarce minerals and create a resilient and ethical supply chain, comes with many challenges. The design of an effective and high-performing battery management system (BMS) for such technology is one of those challenges. In this work, a physics-based model describing the two-phase transition operation of an iron-phosphate positive electrode—in a graphite anode battery—is integrated with a machine-learning model to capture the hysteresis and path-dependent behavior during transient operation. The machine-learning component of the proposed “hybrid” model is built upon the knowledge of the electrochemical internal states of the battery during charge and discharge operation over several driving profiles. The hybrid model is experimentally validated over 15 h of driving, and it is shown that the machine-learning component is responsible for a small percentage of the total battery behavior (i.e., it compensates for voltage hysteresis). The proposed modeling strategy can be used for battery performance analysis, synthetic data generation, and the development of reduced-order models for BMS design.

摘要随着对减少电池制造过程中稀有矿物的依赖及创建有韧性和道德供应链的需求增加，采用磷酸锂电池面临着诸多挑战。针对这种技术，设计一个有效且高性能的电池管理系统（BMS）是这些挑战之一。在这项工作中，一个基于物理的模型被用来描述石墨阳极电池中磷酸铁正极的两相变操作，该模型与机器学习模型相结合，以捕捉瞬态操作中的滞后和路径依赖行为。我们提出的“混合”模型的机器学习部分，建立在对电池在多个驱动模式下充放电过程中电化学内部状态的了解之上。该混合模型经过15小时的驱动实验验证，结果表明，机器学习部分仅占电池总行为的一小部分（即，它补偿了电压滞后）。本工作提出的模型策略可用于电池性能分析、合成数据生成以及开发用于BMS设计的降阶模型。