Reliability enhancement of state of health assessment model of lithium-ion battery considering the uncertainty with quantile distribution of deep features
针对现有数据驱动模型忽略电池健康预测不确定性的问题，提出基于深度学习的健康状态评估模型，通过深度特征的分位数分布给出带置信区间的评估结果，并开发Wasserstein距离分位数Huber损失函数优化模型，在NASA数据集上验证了有效性和可靠性。
Lithium-ion batteries (LIBs) are widely used in many fields, such as electric vehicles and energy storage, and directly impact the device performance and safety. Therefore, the state of health (SOH) assessment is critical for LIB usage. However, most of the existing data-driven SOH modeling methods overlook the inherent uncertainty in battery health prediction, which decreases the reliability of the model. To address this issue, this paper proposes a novel SOH assessment model based on the deep learning framework. The SOH results are derived from the quantile distribution of deep features, giving the SOH values with associated confidence intervals. This enhances the reliability and generalization of SOH assessment results. Additionally, to complete the optimization of the deep model, a Wasserstein distance-based quantile Huber (QH) loss function is developed. This function integrates Huber loss and quantile regression loss, enabling the model to be optimized based on a distribution output. The proposed method is validated using the NASA dataset, and the results confirm that the proposed method can effectively estimate the SOH of LIB while accounting for uncertainty. The incorporation of SOH distribution enhances the reliability and generalization ability of the SOH assessment model.