Methods for estimating the accumulated nitrogen concentration in anode of proton exchange membrane fuel cell stacks based on back propagation neural network

Excellent performance is an important indicator to ensure the commercial development of fuel cells. During the operation of a fuel cell system, nitrogen accumulation will occur at the anode of the fuel cell stack because of hydrogen circulation and nitrogen permeation across the membrane. Nitrogen accumulation will lead to the degradation of fuel cell performance. A suitable purge strategy can effectively reduce the excessive nitrogen concentration. The formulation of the purge strategy is closely related to the change in nitrogen concentration, so the accurate estimation of nitrogen concentration is quite important. In this paper, the effect of nitrogen concentration on fuel cell performance under different current densities is studied, and the anode nitrogen concentration in the fuel cell stack is estimated by back propagation neural network. The identification result of nitrogen concentration based on the voltage of every single cell is more accurate. Without considering aging and working condition changes, mean absolute errors of estimated results are 0.75%, 0.67%, 0.62%, and 0.73%, and root mean square errors are 1.09%, 0.97%, 0.88%, and 0.99% at different current densities of 0.6, 0.8, 1.0, and 1.2 A·cm−2, respectively. The results indicate that the estimation method of nitrogen concentration for fuel cell anode based on back propagation neural network has a high accuracy, which can provide a new method for formulating purge strategy for fuel cell system.