Sensitivity analysis of a CNN–LSTM prognostic framework for proton exchange membrane fuel cells: Effects of sliding window size and training data allocation

Abstract

For proton exchange membrane fuel cell (PEMFC) prognostics, deploying deep learning models in real applications depends not only on the network architecture but also on carefully chosen hyperparameters and training strategies. Hybrid Convolutional Neural Network–Long Short-Term Memory (CNN–LSTM) models can attain high predictive accuracy, yet their sensitivity to practical implementation choices has not been systematically quantified. This work addresses that gap by performing a methodological assessment of a representative CNN–LSTM framework rather than proposing a new architecture. Using the static-load IEEE PHM 2014 FC1 dataset as a controlled benchmark, we examine how two key factors—sliding window length and training data partitioning—jointly affect short-term accuracy and long-term forecast stability. Our results show that the hybrid CNN–LSTM reduces the root mean square error (RMSE) of short-term voltage prediction by 49% compared with a standalone LSTM baseline. For recursive long-horizon lifetime forecasting, a moderately sized sliding window of 40 h is identified as optimal, keeping the remaining useful life prediction error within ±5 h. In addition, the model exhibits strong robustness with respect to training set size: once more than 500 h of data are used for training, the variation in RMSE remains below 0.001 V. By confining the analysis to a static-load test case, we isolate the influence of these implementation parameters and provide practical, data-driven guidance for configuring CNN–LSTM-based prognostic models in PEMFC health management. The proposed validation methodology can be extended in future work to dynamic-load scenarios.

1. Introduction

The growing concerns over greenhouse gas emissions and fossil fuel depletion have accelerated the search for clean, efficient, and sustainable energy solutions [1–2]. In parallel, substantial progress has been made in broader energy conversion and storage technologies (e.g., electrocatalysis and advanced functional materials), which underpin low-carbon energy systems [3–6]. Proton exchange membrane fuel cells (PEMFCs), a key hydrogen-based technology, offer high efficiency, low operating temperatures, rapid response times, and zero emissions, making them a promising option for transportation, stationary power generation, and portable electronic applications. However, the widespread adoption of PEMFCs is hindered not by fundamental technological limitation, but primarily by challenges related to performance degradation, high system costs, and, crucially, limited durability [7].

The membrane electrode assembly (MEA), comprising the proton exchange membrane, catalyst layer, and gas diffusion layer, is prone to degradation during operation, leading to a decline in efficiency and lifespan. Primary degradation mechanisms include chemical breakdown, mechanical wear, thermal stress, and contamination. Chemical breakdown, as detailed by Choi et al. [8] focusing on membrane design and Kamiya et al. [9] on proton transfer mechanisms, presents fundamental material challenges, while Kreuer [10] highlights the overall complexity of proton conduction. Seo et al. [11] and Tang et al. [12] investigated physical MEA degradation; however, their focus remained on specific operational stresses (on/off cycles and membrane degradation, respectively) rather than comprehensive predictive modeling. Virkar and Zhou [13] developed a kinetic model for catalyst degradation, offering valuable mechanistic insights, but lacking the adaptability of data-driven approaches for real-world, dynamic conditions. Catalyst degradation also involves Pt particle aggregation [13], migration [14], and detachment [15], reducing the active surface area and contributing to performance decline.

Effective lifetime prediction is essential for optimizing maintenance strategies, improving system reliability, and enhancing commercial viability. Prognostics and health management (PHM) technology [16] offers the potential for extending PEMFC service life and enabling systematic maintenance [17–18]. While PHM encompasses both fault diagnosis and remaining useful life (RUL) prediction, this work focuses specifically on the latter, recognizing its critical role in proactive maintenance planning and resource allocation [19]. Accurate RUL prediction allows for early detection of degradation, enabling preventative measures and optimizing fuel cell replacement schedules.