Multi objective optimization using artificial neural network to maximize the power output of PEMFCs

Designing a PEM fuel cell model is exceedingly challenging because of its multivariate in nature. Optimization is required to achieve highest operating condition. Neural Network Model is one of the possible methods to solve complex problems. The polarisation curve of a PEMFC (Proton Exchange Membrane Fuel Cell) is investigated in this paper in relation to the effects of seven parameters, including temperature, relative humidity in the cathode, relative humidity in the anode, anode stoichiometry, cathode stoichiometry, partial pressure of H2, and partial pressure of O2, using an ANN (artificial neural network) model. Where model geometric parameters i.e. Channel width, Channel depth, Channel length, Rib width, Cell width, GDL thickness, CL thickness, Membrane thickness of PEMFC was constant. Initially single Objective Function (Output Power) is predicted. The research presented here makes predictions about a PEMFC stack's electrical performance under multiple operating conditions. Mathematical model was further verified using laboratory data. Co-efficient of Determination (R2), Mean Square Error (MSE), and Mean Absolute Error (MAE) was determined using the fuel cell stack voltage model and stack power model. The model results show the possibility of using ANN in the implementation of such models to predict the PEMFC system's steady-state behaviour. Highlights. Experimentally data useful for investigation and work on PEMFCs. ANN models to predict the steady state behaviour of the PEMFC system for different operating conditions. Single Objective Function (Output Power Predicted). ANN-Multi Objective function is presented to predict Efficiency and output Power simultaneously. ANN-MOO model is Validated using Laboratory Dat