Application of Neural Network approach for Proton Exchange Membrane fuel cell systems

Abstract

Artificial Intelligence (AI) techniques, particularly the Neural Networks (NNs), are recently having significant impact on power electronics. In a Proton Exchange Membrane (PEM) fuel cell system, there is a strong relationship between the available electrical power and the actual operating conditions: gas conditioning, membrane hydration state, temperature, current set point, etc. Thus, a 'minimal behavioural model' of a fuel cell system able to evaluate the output variables and their variations is highly interesting. In this paper, we are interested in controlling the powers by using Neural Networks Controllers under the assumption that any system of production is subjected permanently to load steps change variations. So a static production system including a Proton Exchange Membrane Fuel Cell (PEMFC) is subjected to variations of active and reactive power. Then, the goal is to make the system follows these imposed variations. In this work, a PEM fuel cell NN model is proposed using a quasi-Newton method and implemented on Matlab/Simulink® software, Levenberg-Marquardt training algorithm, activation functions and their causes on the effectiveness of the performance modelling are discussed, the quasi-Newton NN control is described and results from the analysis as well as the limitations of the approach are presented.