A proton exchange membrane fuel cell degradation prediction method fusing crossformer and physical constraints
By employing a three-stage training strategy that combines one-dimensional convolution enhancement and Crossformer with electrochemical physical constraints, the problems of insufficient multi-dimensional feature extraction and poor physical interpretability in proton exchange membrane fuel cell degradation prediction are solved, achieving high-precision, stable multi-condition adaptability and degradation prediction under small sample conditions.
CN122412907APending Publication Date: 2026-07-17UNIV OF SCI & TECH OF CHINA
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- UNIV OF SCI & TECH OF CHINA
- Filing Date
- 2026-04-14
- Publication Date
- 2026-07-17
Smart Images
Figure CN122412907A_ABST
Abstract
本发明公开了一种融合Crossformer与物理约束的质子交换膜燃料电池退化预测方法，属于新能源电池健康管理技术领域。本发明针对现有燃料电池退化预测方法存在的多维度时序特征提取不足、物理可解释性差、小样本泛化能力弱、多工况适应性差等问题，构建卷积增强的Crossformer混合预测架构，将质子交换膜燃料电池电化学老化机理嵌入神经网络损失函数，形成数据拟合与物理约束相结合的复合损失函数，并采用三阶段渐进式训练策略实现模型协同优化；本发明能够同时捕获多变量时间序列的局部波动特征、长期时序依赖与跨维度耦合关系，在物理规律约束下提升小样本与多工况下的预测精度、泛化能力与物理一致性。本发明适用于质子交换膜燃料电池健康状态评估、寿命预测与系统健康管理，可广泛应用于车载燃料电池、分布式发电与便携式电源等场景。
Need to check novelty before this filing date? Find Prior Art