燃料电池寿命混合预测方法综述OA

Fuel cells find extensive application across multiple sectors due to their high conversion effi-ciency,operational stability,and environmental friendliness.However,their performance gradually degrades with use,making the accurate prediction of their remaining useful life(RUL)crucial for maintenance and lifespan extension.Hybrid prediction methods represent a significant approach in fuel cell lifespan forecasting,combining the strengths of physical models and data-driven models.This paper first briefly introduces single model-driven and data-driven RUL prediction methods.It then focuses on analysing the primary hybridisation strategies within hybrid prediction methods,dis-cussing the advantages and disadvantages of different approaches.By reviewing existing research on mainstream hybridisation strategies—model-data and data-data hybridisation—the study finds both approaches possess distinct strengths:model-data hybridisation enhances prediction accuracy and generalisation capability,yet incurs higher computational complexity due to its intricate framework;while data-data hybrid prediction simplifies modelling and improves computational efficiency while maintaining good prediction accuracy,it suffers from high data dependency and poor physical inter-pretability.Finally,the paper summarises the advantages of hybrid prediction methods in enhancing RUL prediction accuracy and reliability,along with their development trends,thereby providing ro-bust support for RUL research in fuel cells.