As lithium-ion batteries are gaining more attention in various industries, analysis of their performances especially state-of-charge becomes critical. Due to manufacturing tolerance, battery model parameters may deviate from the nominal parameters. Such a parameter mismatching issue could result in an inaccurate cell state-of-charge estimation. Conducting cell characterization test for each cell, however, is time consuming and economically not possible in reality. This paper proposes a probabilistic approach for battery state-of-charge estimation considering the cell-to-cell variability so that reliable state-of-charge estimation can be obtained for a batch of similar cells. Therefore, parameter characterization test for each cell is not required while ensuring the state-of-charge estimation reliability. The proposed method consists of two major technical components. Firstly, battery model parameters are calibrated and modeled as Gaussian processes over the state-of-charge domain to account for the cell-to-cell variability. Secondly, state-of-charge variability under any charging/discharging profile is effectively quantified through the seamless integration of the extended Kalman filter and an uncertainty quantification method. As such, confidence-based state-of-charge estimation can be produced. Accurate and efficient estimation of the state-of-charge uncertainty are demonstrated for four battery loading profiles.

随着锂离子电池在各个行业中越来越受到关注，对其性能尤其是充电状态的分析变得至关重要。由于制造公差，电池模型参数可能会偏离标称参数。这种参数不匹配问题可能导致电池充电状态估计不准确。然而，对每个电池进行电池特性测试既费时又经济，实际上是不可能的。本文提出了一种考虑电池间可变性的电池充电状态估计的概率方法，以便可以为一批类似的电池获得可靠的充电状态估计。因此，在保证荷电状态估计可靠性的同时，不需要对每个电池进行参数表征测试。所提出的方法由两个主要技术组成部分组成。首先，电池模型参数被校准和建模为荷电状态域上的高斯过程，以解释电池间的可变性。其次，通过扩展卡尔曼滤波器和不确定性量化方法的无缝集成，可以有效量化任何充电/放电曲线下的充电状态变化。因此，可以产生基于置信度的荷电状态估计。对四种电池负载曲线展示了对充电状态不确定性的准确有效估计。通过扩展卡尔曼滤波器和不确定性量化方法的无缝集成，可以有效量化任何充电/放电曲线下的充电状态变化。因此，可以产生基于置信度的荷电状态估计。对四种电池负载曲线展示了对充电状态不确定性的准确有效估计。通过扩展卡尔曼滤波器和不确定性量化方法的无缝集成，可以有效量化任何充电/放电曲线下的充电状态变化。因此，可以产生基于置信度的荷电状态估计。对四种电池负载曲线展示了对充电状态不确定性的准确有效估计。