Abstract
This study develops an advanced electrochemical model integrated with the Teaching-Learning Based Collective Intelligence (TLBCI) algorithm to investigate degradation mechanisms in solid oxide fuel cells (SOFCs), with a focused analysis on nickel (Ni) agglomeration/oxidation at the anode and yttria stabilized zirconia (YSZ) agglomeration at the cathode. Key model parameters are directly extracted from experimental data, enabling accurate performance prediction. The model systematically evaluates the impact of temperature fluctuations on long-term SOFC degradation. Compared to conventional methods (Kalman filters, particle filters) and data-driven approaches (Long Short-Term Memory networks (LSTM), Echo State Networks (ESN)), the proposed mechanism-based model achieves superior accuracy, lower Mean Squared Error (MSE), and enhanced predictive capability in both short- and long-term forecasts. Furthermore, the work provides an in-depth analysis of electrochemical performance decay, including the evolution of overpotential components and material properties. This comprehensive degradation framework advances the understanding of SOFC longevity and provides a theoretical foundation for optimizing cell design, improving reliability, and enhancing operational efficiency—thereby supporting their commercial and industrial deployment (e.g., in distributed generation and backup power systems). The findings offer critical insights for boosting SOFC performance under real-world operating conditions.
•Novel TLBCI-model predicts SOFC decay from Ni/YSZ agglomeration and oxidation.•TLBCI algorithm surpasses Kalman/LSTM in short & long-term forecast accuracy.•Decouples degradation mechanisms, revealing overpotential and material evolution.•Quantifies joint acceleration of Ni decay by temperature and current.•Rapid thermal cycling & startups significantly hasten performance decay.