Impedance model for diagnosis of water management in fuel cells using artificial neural networks methodology

The objective of this work is to define and to implement a simple method to assess the impacts of relative humidity and operating time on the fuel cell impedance. The method is based on the physical model of Randles with CPE and a mathematical tool for identifying various parameters based on the least squares method. The objective of the theoretical model development is the model implementation of the control system and water management of predictive diagnostics. Artificial neural networks are used to create the optimum impedance model. The model is applied for the identification of all resistors (internal resistors measured at high frequency, biasing resistors measured at high frequency) which are characterized by a high sensitivity for both cases, the flooding or drying of the cell heart (membrane and electrodes). This model is able to easily generate Nyquist diagram for any condition of relative humidity and operating time, it helped define the stack hydration status. Based on the obtained results, the model demonstrated a best flexible response, accurate and fast. The developed model can be integrated into a water management control system in PEM fuel cells.