In the field of fuel cells, early detection of faulty conditions can significantly improve the lifetime. Then, signal analysis techniques such as electrochemical impedance spectroscopy combined with machine learning algorithms can generate a representation of the system state of health space using data in known conditions. Although the onboard measurement of EIS can be done by controlling the harmonic content of the power converter at the output of the fuel cell, the implementation of in-vehicle diagnostic algorithms is still limited by the absence of a large database listing the evolution of performance throughout the life cycle. This paper presents a fast-diagnostic method able to consider the occurrence of new data to adapt the dimensional space representing the health state and compensate for the lack of data. Available measurements come from two low-temperature proton exchange membrane fuel cell technologies characterized by two laboratories. The results presented in the paper show that the automatic parameter selection provides performances as good as the ones obtained by an expert. The feasibility of the approach has also been demonstrated on a low-cost embedded platform.

The goal of increasing efficiency and durability of fuel cells can be achieved through optimal control of their operating conditions. In order to implement such controllers, accurate and computationally efficient fuel cell models must be developed. This work presents a hybrid (physics-based and data-driven), control-oriented model for approximating the output voltage of proton exchange membrane fuel cells (PEMFCs) while operating under dynamical conditions. First, a physics-based model, built from simplified electrochemical, membrane dynamics and mass conservation equations, is developed and validated through experimental data. Second, a data-driven, neural network (echo state network) is trained, fitted and tested with the same dataset. Then, the hybrid model is formed as a parallel structure, where the simplified physics-based model and the trained data-driven model are merged through an algorithm based on Gaussian radial basis functions. The merging algorithm compares the output of both single models and assigns weights for computing the prediction of the hybrid result. The proposed hybrid model structure is successfully trained, validated and tested with an experimental dataset originating from fuel cells within an automotive PEMFC stack. The hybrid model is assessed through the mean square error index, with the result of a low tracking error.