Research on Fault Diagnosis Based on Attention‐Based Bidirectional Long–Short Term Memory Network and Multilayer Perceptron in Multistack Fuel Cell System and Energy Management Strategy of Hybrid Power System

The multistack fuel cell system (MFCS) enhances power output, flexibility, and fault tolerance in fuel cell technology. This study introduces an integratedframework combinig fault diagnosis and energy management strategy (EMS) for MFCS hybrid systems. A high‐fidelity model is developed, followed by fault embedding experiments to create a dataset. Novel diagnostic methods—bidirectional long‐short term memory network and a multilayer perceptron— achieve accuracies of 94.19% and 95.65%, outperforming traditional methods. An efficient EMS using power demand prediction and sequential quadratic programming (SQP), optimizing real‐time energy allocation between the MFCS and battery, adapting dynamically under faults. Validated on a 4 × 25 kW platform, the strategy improves performance in both normal and faulty conditions, enhancing state of charge, tracking accuracy, and energy efficiency. In summary, this study advances MFCS hybrid power systems by 1) establishing accurate simulation models and experimental platforms, 2) developing high‐performance fault diagnosis methods leveraging attention mechanisms, and 3) designing innovative EMS strategies for real‐time energy optimization. These contributions collectively enhance the reliability, efficiency, and operational performance of fuel cell hybrid systems, providing a robust foundation for future applications in sustainable energy systems.