Accurate prediction of the remaining useful life (RUL) of fuel cells is a critical enabler for ensuring reliable, safe, and sustainable operation of hydrogen-powered aviation systems. As the aviation sector moves toward decarbonization through hydrogen-electric propulsion, the ability to monitor and forecast fuel cell degradation becomes essential for predictive maintenance, system optimization, and mission-critical decision-making. However, conventional RUL prediction models often fail to effectively capture the complex spatial and temporal dependencies embedded in degradation signals, resulting in limited accuracy and robustness under dynamic operational conditions.

To address these limitations, this study introduces a novel hybrid deep learning framework WOA-CNN-MMHA that synergistically combines Convolutional Neural Networks (CNN), a Masked Multi-Head Attention (MMHA) mechanism, and the Whale Optimization Algorithm (WOA). The CNN module extracts localized spatial features from high-dimensional sensor data, while MMHA models long-range temporal dependencies and preserves critical sequential aging patterns. WOA is employed to optimize hyperparameters, enhancing convergence speed and reducing reliance on manual tuning. This integrated approach enables the model to simultaneously capture both fine-grained degradation characteristics and global deterioration trends, which are vital for accurate RUL estimation.

This work highlights the importance of adopting a holistic modeling strategy that bridges multi-scale feature extraction with intelligent optimization, paving the way for enhanced prognostics in next-generation hydrogen-electric aircraft. By enabling proactive maintenance and improved energy management, the proposed framework contributes to the broader objective of integrating efficient and implementable hydrogen technologies into sustainable regional aviation.