Abstract:Optimal control of proton exchange membrane fuel cell (PEMFC) stacks is required to establish an accurate stack model. The existing PEMFC stack model based on Elman neural network has good precision, but this model still has problems such as easy to fall into local extremum and the result cannot be reproduced. Considering introducing adaptive Levy flight and preference random walk into the basic Krill Herd (KH) algorithm, an improved Krill Herd (IKH) algorithm is obtained to optimize the initial parameters of the neural network, and then a PEMFC stack model based on IKH-Elman network is established.The simulation results show that the IKH algorithm used to optimize the neural network can guarantee higher search accuracy and faster convergence