Correlation analysis and prediction of PEM fuel cell voltage during start-stop operation based on real-world driving data

The accurate prediction of fuel cell voltage is critical for fuel cell health management, which directly affects the performance and cost of fuel cell vehicles. This paper proposes analysis and prediction approaches for the start-stop voltage of fuel cell based on data from a proton exchange membrane fuel cell (PEMFC) bus operated under real traffic conditions in Chengdu, China. First, three correlative analysis methods are used to identify the correlation between the PEMFC voltage and other variables during start-stop operation. Then, the original sequence of PEMFC voltage at start-stop is processed using the grey model, residual grey model, logarithmic-power function grey model, and cotangent function grey model. Four grey prediction models are combined and optimized based on a backpropagation neural network (BPNN). Finally, an organic grey BPNN model (OGNNM) is established to predict the PEMFC voltage during start-stop operation and analyze its accuracy. Results indicate that the PEMFC voltage changes sharply during start-stop, which is closely connected to the current, hydrogen gas tank pressure, and PEMFC stack temperature. In addition, the proposed OGNNM has higher prediction precision than the other data-driven models that provides a basis for the control of fuel cell hybrid systems.