Optimal estimation of the PEM fuel cells applying deep belief network optimized by improved archimedes optimization algorithm

The present study proposes a new efficient methodology for optimal model identification of the Proton-exchange membrane fuel cell (PEMFC) stacks based on an improved version of a Deep Belief Network (DBN). The proposed DBN has been updated by a new metaheuristic to provide the minimum relative error between the experimental output voltage and the network output data during simulation of the nonlinear transient behavior of the Proton-exchange Membrane Fuel Cells (PEMFC). To develop the effectiveness of the DBN, an improved version of the Archimedes optimization algorithm (IAOA) has been developed. The results of training and testing of the proposed method are compared with the original DBN model to indicate the method's effectiveness. Simulations showed 34.0879 and 28.5016 V for the DBN and the suggested DBN-IAOA methods, respectively. This indicates the higher performance of the suggested method toward the original DBN model and its well-organization for modeling the PEMFC stacks. •A new method for optimal model identification of the Proton-exchange membrane fuel cell.•An improved version of a Deep Belief Network (DBN) is used for model identification.•The DBN is developed by an improved version of Archimedes optimization algorithm.•The idea is to minimize the relative error between the experimental and the PEMFC model voltage.