The commercialization of Proton Exchange Membrane Fuel Cells (PEMFC) is constrained by high costs and short lifespans. Conducting degradation prediction research will be beneficial for enhancing the durability and extending the lifespan of PEMFC. In this article, a novel long short-term memory (LSTM) model combined with the quantum-behaved particle Swarm optimization (QPSO) is proposed to achieve high accuracy of PEMFC degradation prediction. LSTM is utilized to describe the voltage degradation of PEMFCs, and QPSO is utilized to enhance the prediction performance by optimizing the hyperparameters within LSTM. The effectiveness of the presented prediction method was validated utilizing the durability test data of a fuel cell stack under constant current load. Experimental results demonstrate that presented method shows great prediction accuracy of PEMFC aging, and its prediction accuracy significantly surpasses conventional methods.