Modelling and Analysis of Polymer Electrolyte Membrane Fuel Cell using Artificial Neural Networks

Abstract

To avoid extensive and costly experiments, the fuel cells developers use detailed newlinecell and stack models for economic ssessments and development purposes. From the results of the Pure Mathematical model simulations, conducted for a broad range of operating conditions, performance charts can be constructed. However, these models are rather detailed descriptions of the physical newlineprocesses occurring in the fuel cells and hence they are intricately complex and newlinecumbersome, especially in operating point analysis and optimization. In the proposed work, an alternative approach to mathematical models based on statistical data-driven artificial neural networks (ANNs) is introduced. Applications of ANNs include a large variety of engineering applications like pattern recognition (protein analysis, spectroscopy and fingerprint identification), as well as behavior prediction and function approximation (stock market forecasting, energy demand forecasting and process control newlinesystems). All these methods are analogous to the central nervous system, including some key features such as parallel information processing and high connectivity similar to the human brain. But ANNs are not programmed. ANNs undergo training to learn by experience and based on past data. In this work, a newlinetwo-layer feed-forward network will be trained with the back propagation algorithm to learn the performance parameters in a planar PEM Fuel Cell. The data used during the training will be generated with a validated physical model and simulation model. In future applications, the data will be available from 5 various fuel cells experiments. However, this tool can be used to design fuel cells with various advantages such as reduced calculation time, economical newlineduring experimental phase and manufacturing process of these units. The first part of the proposed work deals with a general overview of the chosen method, while the next part of the proposed work reports the results of the ANN model applied to the PEM Fuel Cells.