Proton Exchange Membrane (PEM) water electrolysis system is one of the promising technologies to produce green hydrogen from renewable energy sources (wind and solar). However, performance and dynamic analysis of PEM water electrolysis systems are challenging due to the intermittent nature of such sources and involved multi-physical behaviour of the components and subsystems. This study proposes a generic dynamical model of the PEM electrolysis system represented in a modular fashion using Bond Graph (BG) as a unified modelling approach. Causal and functional properties of the BG facilitate the formal PEM electrolyser model to adapt and to fit the different configurations of the electrolyser ranging from laboratory scale to industrial scale. The system-specific key parameter values are identified optimally for a laboratory-scale electrolyser system running on a multi-source energy platform using experimental data. The mean absolute percentage error between simulation and experimental data is found to be less than 5%. The performance characteristic curves of the electrolyser are predicted at different operating temperatures using the identified key parameters. The predicted performance is in good agreement with the expected behaviour of the electrolyser found in the literature. The model also estimates the different energy losses and the real-time efficiency of the system under dynamic inputs. With these capabilities, the developed model provides an economical mean for design, control, and diagnosis development of such systems.

The multimodel approach was recently developed to deal with the issues of complex processes modeling and control. Despite its success in different fields, it still faced with some design problems, and in particular the determination of the models and of the adequate method of validities computation. In this paper, we propose a neural approach to derive different models describing the process in different operating conditions. The implementation of this approach requires two main steps. The first step consists in exciting the system with a rich (e.g. pseudo random) signal and collecting measurements. These measurements are classified by using an adequate Kohonen self-organizing neural network. The second step is a parametric identification of the base-models by using the classification results for order and parameters estimation. The suggested approach is implemented and tested with two processes and compared to the classical modeling approach. The obtained results turn out to be satisfactory and show a good precision. These also allow to draw some interpretations about the adequate validities’ calculation method based on classification results.