Noise level estimation and classification in hydrogen fuel cell-based distributed energy systems: A comprehensive analysis

Abstract

Hydrogen fuel cells (FCs) are a critical clean energy technology significantly contributing to carbon emission reduction and environmental sustainability. However, fuel cell-based distributed energy systems (FC-DESs) are highly susceptible to noise disturbances, adversely affecting their stability and efficiency. To overcome this challenge, a robust algorithm is proposed to monitor and classify noise levels in FC-DESs under static and dynamic noise conditions. The algorithm consists of four stages: signal measurement, pre-processing, signal processing, and decision-making. By employing discrete wavelet transform (DWT)-based denoising techniques, the algorithm processes noisy signals effectively, improving overall signal quality. In the decision-making stage, a classification approach leveraging the Heaviside step function is employed to classify signal-to-noise ratio (SNR) values into distinct noise levels. Performance evaluations reveal that the algorithm achieves classification accuracies of 100% under static and 95% under dynamic conditions, as validated through confusion matrix analysis. Furthermore, the algorithm achieved a percentage error as low as 0.69% and an R2 value of 0.9998 under optimal configurations, indicating its noise-level estimation precision. These findings demonstrate the algorithm's effectiveness in enhancing FC-DES systems' stability and operational reliability, thereby contributing to more efficient energy management under variable operating conditions.