Accurate prediction of fuel cell degradation characteristics is crucial for their control and health diagnostics. A physics-informed neural network was constructed in this study, where high-dimensional implicit physical equations related to voltage were established through a Long Short-Term Memory (LSTM) network architecture. The voltage decay dynamics over time in fuel cell operation were simulated using a deep neural network (DNN) to characterize the electrochemical kinetics. The proposed method was demonstrated to achieve superior prediction accuracy while producing degradation patterns that consistently align with actual Proton Exchange Membrane Fuel Cell (PEMFC) aging mechanisms.