Owing to load variations, the power tracking of fuel cells in automotive applications is emphasized. We address the control strategies to meet varying power demands while optimizing system efficiency. Based on the control‐oriented proton exchange membrane fuel cell model, the relative gain array is analyzed to determine variable parings in control design. As the dynamics of two control variables are different, a cascade control architecture is adopted. A feedforward‐feedback composite control is used to achieve fast response and eliminate the stable error. To avoid oxygen starvation at power transients, an oxygen excess ratio constrained strategy is proposed to limit the load current if there is lack of reaction oxygen. A mixed‐sensitivity synthesis is applied to suppress parameter perturbation and improve system robustness. To further suppress the fluctuations of the oxygen excess ratio and achieve the maximum system efficiency, an oxygen excess ratio invariant strategy is developed, which coordinates load current with airflow. A comparative study is conducted, with two scenarios of power tracking and fuel cell degradation. The results show that the mixed‐sensitivity strategy has a fast response in power tracking and effective perturbation suppression performance, while the oxygen excess ratio invariant strategy maximizes the system efficiency.