Due to their clean reaction product–water, fuel cells (FCs) are increasingly used as power sources for vehicles and ships. However, they are prone to faults during typical degradation conditions like vehicle start-stop and variable loads, some of which can cause irreversible damage. Early fault diagnosis is crucial for extending FCs’ lifespan. Current research often involves extensive fault experiments and accelerated life tests to create private datasets. This leads to excessive hydrogen consumption and increased carbon emissions, contradicting the goal of clean energy. This study adopts a data mining method, focusing on electrochemical impedance spectroscopy (EIS). It explores the EIS evolution patterns of FCs under different states, uses these patterns for data filling, and minimizes external disturbance from EIS. By dynamically expanding real-time impedance data for a lifecycle phase and generating a novel spatial map of impedance over time, it achieves fault diagnosis.