Fault diagnosis of the hybrid system composed of high-power PEMFCs and ammonia-hydrogen fueled internal combustion engines using ensemble deep learning methods

This study proposes an ensemble deep learning method for diagnosing faults in a hybrid power generation system composed of high-power proton exchange membrane fuel cells (PEMFCs) and ammonia-hydrogen fueled internal combustion engines (AHICEs). The ensemble method integrates principal component analysis, a multi-scale convolutional neural network, and an optimized support vector machine to evaluate the system's reliability and accuracy under various single and multiple fault conditions. The results demonstrate that the ensemble deep learning method achieves an overall fault diagnosis accuracy of 98.15% for PEMFCs when diagnosing both single and multiple faults. Similarly, for the internal combustion engine, the method attains an accuracy of 97.85% under equivalent fault conditions. Additionally, when the hybrid system simultaneously encounters multiple faults in both components, the ensemble method achieves a fault diagnosis accuracy of 96.67%. Compared to traditional methods such as support vector machine, backpropagation neural networks, and principal component analysis-support vector machine models, the proposed approach consistently demonstrates superior diagnostic performance. By leveraging its inherent advantages such as dimensionality reduction, enhanced feature extraction, and robust classification, the proposed method provides an efficient and accurate fault diagnosis in hybrid power generation systems.