Parametric analysis of proton exchange membrane fuel cell performance by using the Taguchi method and a neural network

This study proposes a novel parameter optimization method, capable of integrating the neural network and the Taguchi method for parametric analysis of proton exchange membrane fuel cell (PEMFC) performance. Numerous parameters affecting the PEMFC performance are analyzed, such as fuel cell operating temperatures, cathode and anode humidification temperatures, operating pressures, and reactant flow rate. In the traditional design of experiments, the Taguchi method has been popularly utilized in engineering. However, the parameter levels selected to form the orthogonal array in the Taguchi method are discrete, preventing the estimation of the real optimum. This study used the Taguchi method to acquire the primary optimums of the operating parameters in the PEMFC. Each row in the orthogonal array together with its relative responses was used to establish a set of training patterns (input/target pair) to the neural network. The neural network can then construct relationships between the control factors and responses in the PEMFC. The actual optimums of the operating parameters in the PEMFC were obtained by the trained neural network. Experimental results are presented for identifying the proposed approach, which is useful in improving performance for PEMFC and developing electrical system on advanced vehicles and ships.